[go: up one dir, main page]

TW200921679A - Non-volatile memory and method for improved sensing having bit-line lockout control - Google Patents

Non-volatile memory and method for improved sensing having bit-line lockout control Download PDF

Info

Publication number
TW200921679A
TW200921679A TW97121282A TW97121282A TW200921679A TW 200921679 A TW200921679 A TW 200921679A TW 97121282 A TW97121282 A TW 97121282A TW 97121282 A TW97121282 A TW 97121282A TW 200921679 A TW200921679 A TW 200921679A
Authority
TW
Taiwan
Prior art keywords
sensing
memory
bit line
current
group
Prior art date
Application number
TW97121282A
Other languages
Chinese (zh)
Other versions
TWI380307B (en
Inventor
Nima Mokhlesi
Original Assignee
Sandisk Corp
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Priority claimed from US11/759,898 external-priority patent/US7492640B2/en
Priority claimed from US11/759,909 external-priority patent/US7489553B2/en
Application filed by Sandisk Corp filed Critical Sandisk Corp
Publication of TW200921679A publication Critical patent/TW200921679A/en
Application granted granted Critical
Publication of TWI380307B publication Critical patent/TWI380307B/en

Links

Classifications

    • GPHYSICS
    • G11INFORMATION STORAGE
    • G11CSTATIC STORES
    • G11C11/00Digital stores characterised by the use of particular electric or magnetic storage elements; Storage elements therefor
    • G11C11/56Digital stores characterised by the use of particular electric or magnetic storage elements; Storage elements therefor using storage elements with more than two stable states represented by steps, e.g. of voltage, current, phase, frequency
    • G11C11/5621Digital stores characterised by the use of particular electric or magnetic storage elements; Storage elements therefor using storage elements with more than two stable states represented by steps, e.g. of voltage, current, phase, frequency using charge storage in a floating gate
    • G11C11/5642Sensing or reading circuits; Data output circuits
    • GPHYSICS
    • G11INFORMATION STORAGE
    • G11CSTATIC STORES
    • G11C16/00Erasable programmable read-only memories
    • G11C16/02Erasable programmable read-only memories electrically programmable
    • G11C16/04Erasable programmable read-only memories electrically programmable using variable threshold transistors, e.g. FAMOS
    • G11C16/0483Erasable programmable read-only memories electrically programmable using variable threshold transistors, e.g. FAMOS comprising cells having several storage transistors connected in series
    • GPHYSICS
    • G11INFORMATION STORAGE
    • G11CSTATIC STORES
    • G11C16/00Erasable programmable read-only memories
    • G11C16/02Erasable programmable read-only memories electrically programmable
    • G11C16/06Auxiliary circuits, e.g. for writing into memory
    • G11C16/26Sensing or reading circuits; Data output circuits
    • GPHYSICS
    • G11INFORMATION STORAGE
    • G11CSTATIC STORES
    • G11C2211/00Indexing scheme relating to digital stores characterized by the use of particular electric or magnetic storage elements; Storage elements therefor
    • G11C2211/56Indexing scheme relating to G11C11/56 and sub-groups for features not covered by these groups
    • G11C2211/563Multilevel memory reading aspects
    • G11C2211/5634Reference cells

Landscapes

  • Engineering & Computer Science (AREA)
  • Microelectronics & Electronic Packaging (AREA)
  • Computer Hardware Design (AREA)
  • Read Only Memory (AREA)

Abstract

In sensing a group of cells in a multi-state nonvolatile memory, multiple sensing cycles relative to different demarcation threshold levels are needed to resolve all possible multiple memory states. Each sensing cycle has a sensing pass. It may also include a pre-sensing pass or sub-cycle to identify the cells whose threshold voltages are below the demarcation threshold level currently being sensed relative to. These are higher current cells which can be turned off to achieve power-saving and reduced source bias errors. The cells are turned off by having their associated bit lines locked out to ground. A repeat sensing pass will then produced more accurate results. Circuitry and methods are provided to selectively enable or disable bit-line lockouts and pre-sensing in order to improving performance while ensuring the sensing operation does not consume more than a maximum current level.

Description

200921679 九、發明說明: 【發明所屬之技術領域】 此發明概言之係關於非揮發性半導體記憶體(例如,電 可擦除可程式化唯讀記憶體(EEPROM)及快閃EEPROM), 且具體而s係關於其中對與相對高傳導電流記憶體單元相 關聯之位7L線之鎖定施加控制之感測操作及記憶體。 【先前技術】200921679 IX. Description of the Invention: [Technical Field] The present invention relates to non-volatile semiconductor memory (for example, electrically erasable programmable read only memory (EEPROM) and flash EEPROM), and Specifically, s is a sensing operation and memory in which control is applied to the locking of the 7L line associated with a relatively high conduction current memory cell. [Prior Art]

取近,具有電荷非揮發性儲存能力之固態記憶體九六 以封裝成一形狀因素卡之EEPR〇M及快閃EEpR〇M形式之 固態記憶體,已成為多種行動及手持式器件、特別係資訊 用具及消費者電子產品中之首選儲存器。與亦為固態記憶 體之RAM(隨機存取記憶體)不同,快閃記憶體具有非揮發 性’且即使在電源斷開之後亦能保留其儲存之資料。儘管 成本較高’然而快閃記憶體卻正越來越多地用於大容量儲 存應用中。基於旋轉磁性媒體之習用大容量儲存器⑷Close to, solid-state memory with charge non-volatile storage capacity. The solid-state memory in the form of EEPR〇M and flash EEpR〇M packaged into a shape factor card has become a variety of mobile and handheld devices. The preferred storage in appliances and consumer electronics. Unlike RAM (random access memory), which is also a solid-state memory, flash memory is non-volatile and retains its stored data even after the power is turned off. Despite the higher cost, flash memory is increasingly being used in high-volume storage applications. Conventional large-capacity storage based on rotating magnetic media (4)

如,硬驅動器及軟磁碟)不適於行動及手持式環境。此乃 因磁碟驅動器往往較為笨重’易於發生機械故障,且具有 高延時及高功率要求。該等不期望之屬性使得基於磁碟之 儲存器不適用於大多數行動及可攜式應用。另一方 夬 閃記憶體(無論係歲入式還是以一可移除卡之形式)皆可理 f地適於移動及手持式環境,乃因其小大小、低功率消 耗、南速度及而可靠性特徵。 EEPR〇M及電可程式化唯讀記憶體(epr〇m)係 並將新的㈣^或”經程式化之”至其記憶體單元中= 131985.doc 200921679 揮發性記憶體。兩者皆扃,θ ^ ^ (未連接之電晶體結構中利用一浮動 ^通道[ a極,該浮動閘衫位於—半導體基板尹 :上=在源極與汲極區之間。接著在該浮動間 ,=二㈣閘極。該電晶體之臨限電壓特性受到該 ^閉極上㈣留之電荷量之控制。亦即,對於該浮動閉 體之前施加至控制間極之對應ty接通”該電晶 汲極區之間之傳導。 竭品限)以准許其源極與 該㈣閉極可料_電荷㈣,且心可程式化至—臨 限電μ窗(亦稱為―"傳導窗”)内之任—臨限電壓位準。該 臨限電壓窗之大小由器件之最小及最大臨限位準來定界, 而器件之最小及最大臨限位準又對應於可程式化至該浮動 閘極上之電荷範圍。該臨限窗一般而言相依於記憶體器件 之特性、操作條件及歷史。原則上,該窗内之每一不同的 可解析臨限電壓料範㈣可心標識料元之—明確吃 憶體狀態。在臨限電壓被劃分成兩個不同的區日寺,每—記 隐體單元白將能夠儲存—個位元之資料。類似地,在臨阳 電麗窗被劃分成多於兩個不同的區時,每一記憶體單元二 將月b夠儲存多於一個位元之資料。 在通常兩狀態EEPROM單元中’建立至少一個電流斷點 位準以將傳導窗劃分成兩個區。在藉由施加預定之、固定 電壓讀取-單元時,其源極"及極電流藉由與該斷點位準 (或參考電流IREF)相比較而解析成—記憶體狀態。若所讀 取之電流高於該斷點位準之電流,則確定該單元處於—種 131985.doc 200921679 込輯狀〜、(例如,一 "0"狀態)中。另一方面,若該電流小於 °亥斷點位準之電流,則確定該單元處於另-邏輯狀態(例 如 1狀態)中。因此,此兩狀態單元儲存一個數位資 Λ位7L。常常提供一可在外部程式化之參考電流源作為一 記憶體系統之一料來產±斷點位準電流。 為增加記憶體容量,隨著半導體技術狀態的進步,正以 愈來愈兩的密度製造快閃EEPROM器件。另一種增加儲存 谷1之方法係使每一記憶體單元皆儲存多於兩個狀態。 對於—多狀態或多位準EEPROM記憶體單元而言,傳導 窗藉助多於—個斷點劃分成多於兩個區,以使每一單元皆 月b夠儲存夕於一個位元之資料。因此,—既定eeprom陣 列可儲存之資訊隨著每—單元可儲存之狀態之數目之增加 而曰加。已在美國專利第5,172,338號中描述了具有多狀態 或多位準記憶體單元之EEPROM或快閃EEPROM。 虽—記憶體單元之電晶體通常通過兩種機制中之一種 程式化為一,,已程式化"狀態。在”熱電子注入”中,一施加 至汲極之高電壓跨越基板通道區加速電子。同時,一施加 至控制閘極之高電壓拉動熱電子通過一薄閘極電介質到達 f動問極上。在”隨穿注人"中,相對於基板向控制閉極施 電壓以此方式,將電子自該基板拉至介入浮動閘 極。 ^可藉由許多機制來擦除記憶體器件。對於而 '可藉由藉助紫外線輕射自浮動閘㈣除電荷來大量擦 “乂己隐體。對於EEPR0M而言,可藉由相對於控制問極 131985.doc 200921679 向基板施加一高電壓以促使浮動閘極中之電子隧穿通過一 薄氧化物到達基板通道區(亦即,傅勒_諾德翰隧穿)來電擦 除-記憶體單元。通常’可逐一位元組地擦除eepr〇m: 對於㈣EEPRQM而言,可立刻所有塊或—次—個或多個 塊地來電擦除記憶體,其中一塊可由512個或更多個記憶 體位元組組成。For example, hard drives and floppy disks are not suitable for mobile and handheld environments. This is due to the fact that disk drives tend to be bulky and subject to mechanical failures with high latency and high power requirements. These undesired attributes make disk-based storage unsuitable for most mobile and portable applications. The other flash memory (whether in the form of a detachable or a removable card) is suitable for mobile and handheld environments due to its small size, low power consumption, south speed and reliability. feature. EEPR〇M and electrically programmable read-only memory (epr〇m) and new (4)^ or “stylized” into its memory unit = 131985.doc 200921679 Volatile memory. Both are 扃, θ ^ ^ (the unconnected transistor structure utilizes a floating ^ channel [ a pole, the floating trousers are located - the semiconductor substrate Yin: upper = between the source and the drain region. Then in the The floating room, = two (four) gate. The threshold voltage characteristic of the transistor is controlled by the amount of charge left on the (4) of the closed electrode. That is, the corresponding ty applied to the control electrode before the floating body is turned on" The conduction between the dipole regions of the electro-crystals is limited to permit the source and the (4) closed-cell _charge (four), and the heart can be programmed to - the threshold electric window (also known as "" In the conduction window") - the threshold voltage level. The threshold voltage window is delimited by the minimum and maximum threshold levels of the device, and the minimum and maximum threshold levels of the device correspond to the range of charge that can be programmed onto the floating gate. The threshold window generally depends on the characteristics, operating conditions and history of the memory device. In principle, each of the different analyzable threshold voltages in the window (4) can be used to identify the material element. In the case where the threshold voltage is divided into two different districts, the temple will be able to store data for each bit. Similarly, when the Linyang electric window is divided into more than two different areas, each memory unit 2 can store more than one bit of data for the month b. At least one current breakpoint level is established in a typical two-state EEPROM cell to divide the conductive window into two regions. When a predetermined, fixed voltage read-cell is applied, its source " and the pole current are resolved into a memory state by comparison with the breakpoint level (or reference current IREF). If the current being read is higher than the current at the breakpoint level, then the cell is determined to be in the state of 1985. doc 200921679 (for example, a "0" state). On the other hand, if the current is less than the current at the breakpoint level, then the cell is determined to be in another logic state (e.g., state 1). Therefore, the two state units store a digital bit 7L. A reference current source that can be externally programmed is often provided as a source of a memory system to produce a ±-breakpoint level current. In order to increase the memory capacity, as the state of semiconductor technology advances, flash EEPROM devices are being fabricated at increasingly higher densities. Another way to increase the storage of the valley 1 is to store more than two states per memory unit. For a multi-state or multi-bit quasi-EEPROM memory cell, the conduction window is divided into more than two regions by more than one breakpoint, so that each cell can store data for one bit at a time. Therefore, the information that can be stored by an established eeprom array increases with the number of states that each unit can store. An EEPROM or flash EEPROM having a multi-state or multi-bit memory cell has been described in U.S. Patent No. 5,172,338. Although the transistor of the memory cell is usually programmed into one by one of two mechanisms, it has been programmed to state. In "hot electron injection", a high voltage applied to the drain accelerates electrons across the substrate channel region. At the same time, a high voltage applied to the control gate pulls the hot electrons through a thin gate dielectric to the f-movement pole. In the "follow-through", the voltage is applied to the control closed-cell relative to the substrate in such a manner that electrons are pulled from the substrate to the intervening floating gate. ^ The memory device can be erased by a number of mechanisms. 'You can use a light bulb from the floating gate (four) to remove a large amount of rubbing. For EEPR0M, a high voltage can be applied to the substrate relative to the control electrode 131985.doc 200921679 to cause electrons in the floating gate to tunnel through a thin oxide to reach the substrate channel region (ie, Fule-Nuo Dehan Tunneling) Incoming Call Erase - Memory Unit. Usually, 'eepr〇m can be erased one by one: For (iv) EEPRQM, all blocks or one or more blocks can be erased immediately, one block can be 512 or more memory bits. The tuple is composed.

記憶體器件通常包括一個或多個可安裝於一卡上之記憶 體晶片。每一記憶體晶片皆包括一由周邊電路(例如,解 碼器及擦除、寫入和讀取電路)支援之記憶體單元陣列。 更複雜的記憶體器件藉助一外部記憶體控制器操作,該外 部έ己憶體控制實施智慧及較高位準記憶體操作及介接。 現今,正使用許多在商業上成功的非揮發性固態記憶體 器件。該等記憶體器件可係快閃EEPR〇M或可採用其他類 型之非揮發性記憶體單元。在美國專利第5,〇7〇,〇32、 5,095,344、5,315,541、5,343,063 及 5,661,053、5,313,421 及6,222,762號中給出快閃記憶體之實例及其製造系統及方 法。特定而言’美國專利第5,570,3 15、5,903,495、 6,046,935號中描述了具有NAND串結構之快閃記憶體器 件0 亦自具有一用於儲存電荷之介電層之記憶體單元製造非 揮發性記憶體器件。使用一介電層替代先前所述之導電浮 動閘極元件。已由Eitan等人在"NROM: A Novel Localized Trapping, 2-Bit Nonvolatile Memory Cell," (IEEE ElectronMemory devices typically include one or more memory chips that can be mounted on a card. Each memory chip includes an array of memory cells supported by peripheral circuitry (e.g., decoders and erase, write, and read circuitry). More complex memory devices operate with an external memory controller that controls the implementation of intelligent and higher level memory operations and interfaces. Today, many commercially successful non-volatile solid state memory devices are being used. These memory devices can be flash EEPR 〇 M or other types of non-volatile memory cells can be used. Examples of flash memory and methods of making the same are described in U.S. Patent Nos. 5, 5, 095, 344, 5, 315, 541, 5, 343, 063 and 5, 661, 053, 5, 313, 421 and 6, 222, 762. A flash memory device having a NAND string structure is also described in the 'U.S. Patent Nos. 5,570, 3, 5, 903, 495, 6, 046, 935, which are also non-volatile from a memory cell having a dielectric layer for storing charge. Memory device. A dielectric layer is used in place of the previously described conductive floating gate elements. By Eitan et al. "NROM: A Novel Localized Trapping, 2-Bit Nonvolatile Memory Cell," (IEEE Electron

Device Letters,第 21 卷,第 11 號,2000 年 11 月,第 543- 131985.doc 200921679 545頁)中描述可此利用介電儲存元件之記憶體器件。一 ΟΝΟ介電層跨越源極擴散區與汲極擴散區之間之通道延 伸。一個資料位元之電荷集中於毗鄰於汲極之介電層中, 而另負料位元之電荷則集中於毗鄰於源極之介電層中。 舉例而β,美國專利第5,768,192及6,〇11,725號揭示一種具 有夹在兩個二氧化發層之間之俘獲介電層之非揮發性記憶 體單元。多狀態資料儲存係藉由分開讀取介電層内在空間 上分開之電荷儲存區之二元狀態來實施。 為改良讀取及程式化效&,並行地讀取或程式化一陣列 中之多個電荷儲存元件或記憶體電晶體。因此,一同讀取 或程式化-儲存元件邏輯,Τ,。在現有記憶體架構中,一 列通常含有數個交錯頁或其可構成一個頁。將一同讀取或 程式化一頁之所有記憶體元件。 程式化一記憶體單元頁通常涉及一系列交替程式化/驗 證循環。每—程式化循環皆使該記憶體單元頁㈣一個或 多個程式化電壓脈衝。程式化循環後跟一驗證循環,在驗 證循環中讀回每-單S以確定其是否已正確地程式化該單 元。程式化/驗證循環以增加之程式化電壓位準繼續了直 至該頁中之所有單元皆經程式化驗證。 讀取及驗證操作兩者皆藉由執行其中相對於 定該頁之每一記憶體單元之傳導 、刀1值確 平7L乙得導電流或臨限 多個感測循環來實施。一般而言,若記憶體被劃分= 狀癌’則將至少存在…感職環來解 憶體狀態。在許多實施方案中,每-的兄 項衣皆亦可涉及 131985.doc 200921679 兩個或更多個通過。舉例而言,在該等記憶體單元被緊密 封裝時’鄰近電荷儲存元件之間之互動變得顯著且某此感 測技術涉及感測鄰近字線上之記憶體單元以補償由該等互 動引起之誤差。因此,由於正將更多的記憶體單元高度整 合至一晶片中且正將越來越多的狀態封裝至每—記憶體單 元中以增加容量’故讀取及驗證效能由於所需之最代之數 目而受到極大地影響。A memory device that can utilize a dielectric storage element is described in Device Letters, Vol. 21, No. 11, November 2000, 543- 131985.doc 200921679, page 545). A dielectric layer extends across the channel between the source diffusion region and the drain diffusion region. The charge of one data bit is concentrated in the dielectric layer adjacent to the drain, and the charge of the other negative bit is concentrated in the dielectric layer adjacent to the source. For example, a non-volatile memory cell having a trapping dielectric layer sandwiched between two dioxide layers is disclosed in U.S. Patent Nos. 5,768,192 and 6, the disclosure of which is incorporated herein by reference. Multi-state data storage is implemented by separately reading the binary states of the spatially separated charge storage regions within the dielectric layer. To improve read and program efficiency &, read or program multiple charge storage elements or memory transistors in an array in parallel. Therefore, together read or stylize - store component logic, oh,. In existing memory architectures, a column typically contains a number of interleaved pages or it can constitute a page. All memory elements of a page will be read or programmed together. Styling a memory unit page typically involves a series of alternating stylization/verification cycles. Each stylized loop causes the memory unit page (4) to have one or more programmed voltage pulses. The stylized loop is followed by a validation loop in which each-single S is read back to determine if it has properly programmed the unit. The stylization/verification loop continues with the increased stylized voltage level until all units in the page have been programmatically verified. Both the read and verify operations are performed by performing a plurality of sensing cycles in which the conduction of each memory cell relative to the page, the value of the knife 1 is equalized, or the threshold is applied. In general, if the memory is divided into = cancer, then at least... the sensory ring will be present to resolve the state of the body. In many embodiments, each of the brothers can also be involved in 131985.doc 200921679 two or more passes. For example, the interaction between adjacent charge storage elements becomes significant when the memory cells are tightly packaged and some sensing techniques involve sensing memory cells on adjacent word lines to compensate for such interactions. error. Therefore, since more memory cells are being highly integrated into a single chip and more and more states are being packaged into each memory cell to increase capacity, the read and verify performance is due to the most demanding The number is greatly affected.

因而,一般需要高容量及高效能的非揮發性記憶體。特 定而言,需要具有一具有其中上述缺點被最小化之改良感 測效能之南容量非揮發性記憶體。 【發明内容】 根據本發明之一一般態樣,在正並行地感 元頁時’ -位元線鎖定之情形(將一位元線接地:: 超過一預定之電流位準之記憶體單元 以此方式,將允許為達到電流消耗之—既定預算:::識 別及關閉高電流單元之額外感測子循$,且以—選擇性數 目之感測控㈣極電壓感測將導致經相對應於接通單元 之位兀線關閉’而以其他控制閘極電愿感測將不導致任一 =㈣閉?作。通過應用此技術,感測操作之效能因 至位::循裱之數目並在關閉位元線時減小由於位元線 口而產生之雜訊而得以改良。在此上下文中, =:位元―存在於鄰近全局位元線之間之電 在一較佳實施例中 位疋線鎖定係藉助一能夠將位元線 131985.doc •10- 200921679 拉至接地之下拉電路來實施。該下拉電路包括兩個串聯於 位70線與接地之間之通過閘極。該兩通過閘極來自— 閘極,其中一個通過閘極由一下拉啓用或停用控制信號加 以控制且另一個通過閘極由是否感測到討論中之單元具有 一高於或低於一參考電流之電流加以控制。在停用該下拉 電路時,不官所感測之結果如何位元線皆將不接地。在啓 用該下拉電路時,在所感測之結果係來自一高電流記憶體 單元時位元線將被拉至接地。 在一個感測一多狀態記憶體之實施方案(其牽扯相對於 多個狀態中之每一者對記憶體單元頁實施多個感測通過) 中’僅以預定感測通過實施鎖定與識別為高於預定之電流 位準之I己憶體單元相關聯之位元線之步驟。以此方式,收 穫減小總電流及源極偏壓誤差之位元線鎖定益處,同時減 fe其由於更多子循環而具有較長感測時間及用於使所產生 之雜訊衰退之較長等待時間之負效應。為確保一頁中之高 電流狀態之均勻分佈,該頁較佳經編碼以使資料以相對一 致分佈儲存於所有可能的記憶體狀態中。在一較佳實施例 中’該經編碼之資料表現為僞隨機。 根據本發明之另一態樣,位元線鎖定之減小與不超過一 預定之最大電流之記憶體單元頁中流動之總電流相當。以 此方式’位元線鎖定最小化但在總電流(其係資料相依)將 要超過一預定之電流位準引用。 預疋之電流位準在一個感測一多狀態記憶體之實施方案 (其牽扯相對於多個狀態中之每一者對記憶體單元頁實施 131985.doc 200921679 多個感測通過)中,僅在該記憶體單元頁中流動之總電流 將要超過預疋最大電流時以預定感測通過實施鎖定與識別 為高於預定之電流位準之記憶體單元相關聯之位元線之步 驟。 在一個實施例中,提供一電流監視器以監視記憶體單元 頁中流動之總電流。 在另一實施例中,累積對應於高度導電單元之位元線之 數目且使用此資訊來估計記憶體單元頁中流動之總電流。 由下文對本發明之較佳實施例之描述將理解其額外特徵 及優點,此描述應結合隨附圖式來獲得。 【實施方式】 記憶體系統 圖1至圖9圖解§兑明其中可實施本發明之各種態樣之實例 性記憶體糸統。 圖1 0至圖2 1圖解說明本發明之各種態樣及實施例。 圖1示意性地圖解說明其中可實施本發明之一非揮發性 記憶體晶片之功能塊。記憶體晶片1〇〇包含一二維記憶體 單元陣列200、控制電路210及諸如解碼器、讀取/寫入電 路及多工器等周邊電路。 5己憶體陣列200可藉助字線經由列解碼器〇(分成 23 0A、230B)且藉助位元線經由行解碼器26〇(分成26〇A、 260B)來(亦見圖4及5)定址。讀取/寫入電路27〇(分成 270A、270B)允許將並行地讀取或程式化一記憶體單元 頁。一資料I/O匯流排231耦合至讀取/寫入電路27〇。 131985.doc -12· 200921679 #在-:佳實施例中,—頁係由共享相同字線之一列相鄰 思:單7L構成。在另一實施例中’在—列記憶體單元被 』刀成夕個頁之情形下,提供塊多工器250(分成250A及 議)來將讀取/寫入電路27〇多工至個別頁。舉例而言, 將分別由奇數及偶數行記憶體單元形成之_Η工至該 等讀取/寫入電路。Therefore, high capacity and high performance non-volatile memory are generally required. In particular, it is desirable to have a south volume non-volatile memory having improved sensing performance in which the above disadvantages are minimized. SUMMARY OF THE INVENTION According to one aspect of the present invention, a case where a bit line is locked in a parallel sense (a bit line is grounded: a memory cell exceeding a predetermined current level is In this way, it is allowed to achieve the current consumption - the established budget::: the additional sensing sub-clock for identifying and turning off the high current unit, and with a selectable number of sensing control (four) pole voltage sensing will result in the corresponding Turning on the unit's position and turning off the line and using other control gates will not cause any = (4) closed operation. By applying this technology, the performance of the sensing operation is in place: the number of cycles Improvements in reducing noise due to bit line lines when bit lines are turned off. In this context, =: bits - electricity present between adjacent global bit lines in a preferred embodiment The bit line locking is implemented by a pull-down circuit capable of pulling the bit line 131985.doc •10-200921679 to the ground. The pull-down circuit includes two pass gates connected in series between the line 70 and ground. From the gate – the gate, one of which passes The pole is controlled by a pull enable or disable control signal and the other pass gate is controlled by whether the cell in question has a current above or below a reference current. When the pull down circuit is deactivated, As a result of the unsatisfactory sensing, the bit line will not be grounded. When the pull-down circuit is enabled, the bit line will be pulled to ground when the sensed result is from a high current memory unit. An embodiment of a multi-state memory that involves performing multiple sensing passes on a memory cell page relative to each of a plurality of states. 'Performing only by predetermined sensing by locking and identifying higher than a predetermined current The step of registering the bit line associated with the body unit. In this way, the bit line locking benefit of reducing the total current and source bias error is harvested, while reducing it due to more sub-cycles Longer sensing times and negative effects of longer latency for decay of generated noise. To ensure uniform distribution of high current conditions in a page, the page is preferably coded to make the data relatively consistent The cloth is stored in all possible memory states. In a preferred embodiment, the encoded data appears to be pseudo-random. According to another aspect of the invention, the bit line lock is reduced by no more than a predetermined amount. The maximum current flowing in the memory cell page of the maximum current is equivalent. In this way, the bit line lock is minimized but the total current (which is dependent on the data) will exceed a predetermined current level reference. In an embodiment of sensing a multi-state memory (which involves implementing 131985.doc 200921679 multiple sensing passes with respect to each of a plurality of states), only in the memory cell The step in which the total current flowing in the page will exceed the pre-maximum current by a predetermined sense by performing a lock on the bit line associated with the memory cell identified as being above a predetermined current level. In one embodiment, a current monitor is provided to monitor the total current flowing in the memory cell page. In another embodiment, the number of bit lines corresponding to the highly conductive cells is accumulated and this information is used to estimate the total current flowing in the memory cell page. Additional features and advantages will be understood from the following description of the preferred embodiments of the invention. [Embodiment] Memory System Figs. 1 through 9 illustrate the exemplification of an exemplary memory system in which various aspects of the present invention can be implemented. Figures 10 through 21 illustrate various aspects and embodiments of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS Figure 1 is a schematic illustration of functional blocks in which a non-volatile memory wafer of the present invention may be implemented. The memory chip 1A includes a two-dimensional memory cell array 200, a control circuit 210, and peripheral circuits such as a decoder, a read/write circuit, and a multiplexer. The 5 memory array 200 can be via a column decoder via a word line (divided into 23 0A, 230B) and via a bit line via a row decoder 26 (divided into 26 A, 260B) (see also Figures 4 and 5). Addressing. The read/write circuit 27 (divided into 270A, 270B) allows a memory cell page to be read or programmed in parallel. A data I/O bus 231 is coupled to the read/write circuit 27A. 131985.doc -12· 200921679 #在-: In the preferred embodiment, the page is composed of one column sharing the same word line adjacent to each other: a single 7L. In another embodiment, in the case where the 'in-array memory cell' is smashed into a page, the block multiplexer 250 is provided (divided into 250A and discussed) to multiplex the read/write circuit 27 to individual page. For example, the singular and even row memory cells are respectively formed into the read/write circuits.

圖1圖解說明一其中以一對稱形式在記憶體陣列200之相 對側上實施由各種周邊電料料_行存取以使每一側 ^之存取線及電路之密度減半之較佳佈置。因&,列解碼 器分成列解碼器2鳩及2遞,且行解碼器分成行解碼器 260A及細。在其中—列記憶體單元劃分成多個頁之實 細例中’ !乡工器250分成頁多工器25〇八及2遍。類似 地,讀取/寫入電路270分成自底部連接至位元線之讀取、/寫 入電路270A及自頂部連接至位元線之讀取/寫入電路 270B。以此方式,該等讀取/寫入模組之密度且因而感測 模組380之密度實質上減半。 控制電路110係一晶片上控制器,其與讀取/寫入電路 270協作以在記憶體陣列200上實施記憶體操作。控制電路 110通常包含一狀態機112及其他電路(例如,一晶片上位 址解碼器及一功率控制模組(未明確顯示))。狀態機ιΐ2對 記憶體操作提供晶片級控制。該控制電路經由一外部記憶 體控制器與一主機進行通信。 記憶體陣列200通常組織為一沿列及行佈置並可由字線 及位元線定址之兩維記憶體單元陣列。該陣列可根據一 131985.doc 13 200921679 nor類型或一 NANr^M型架構形成。 圖2不意性地圖解說明—非揮發性記憶體單元。記憶體 單元10可由一具有一電荷儲存單元2〇(例如,一浮動閘極 或一介電層)之場效電晶體構建。記憶體單元1〇亦包含一 源極14、一及極16及一控制閘極3〇。 現今正使用許多在商業上成功的非揮發性固態記憶體器 件。該等記憶體器件可採用不同類型之記憶體單元,每一 類型皆具有一個或多個電荷儲存元件。 典型非揮發性記憶體單元包含eepr〇m及快閃 EEPR0M。美國專利第5,595,924號中給出EEpR〇M單元之 實例及其製造方法。美國專利第5,〇7〇 〇32、5 〇95 344、 5,315,541、5,343,063、5,661,053、5,313,421 及 6,222,762 號中給出快閃EEPROM單元之實例、其在記憶體系統中之 使用及其製造方法。特定而言,美國專利第5,57〇,315、 5,903,495及6,046,935號中描述了具單元結構之記 憶體器件之實例。同樣,Eitan等人已在"NR〇M: A Novel Localized Trapping, 2-Bit Nonvolatile Memory Cell, "(IEEE Electron Device Letters,第 21卷,第 U 號,⑼㈧年^ 月, 第543-S45頁)且在美國專利第5,768,192號及第6,〇u,725號 中描述了利用介電儲存元件之記憶體器件之實例。 實際上,通常藉由在向控制閘極施加一參考電壓時感測 跨越一單7L之源電極及漏電極之傳導電流來讀取該單元之 記憶體狀態。因此,對於一單元之浮動閘極上之每—既定 電荷,皆可偵測一相對於一固定參考控制閉極電壓之對應 131985.doc •14· 200921679 =電流。類似地’可程式化至浮動閑極上之電荷之範圍 I疋對應S品限電壓窗或一對應傳導電流窗。 另-選擇為,不在—經劃分之電流窗中偵測傳導電产, 而是可在控制閉極處為一既定記憶體狀態設定臨限電:, 並偵測該傳導電流低於還是高於一臨限電流。在—個實施 方案中,相對於-臨限電流偵測傳導電流係藉由檢查傳導 電流通過位元線之電容放電之速率來完成。 Ο u 圖增對浮動閘極可在任一時刻選擇性地儲存之四個不 同電何Q1-Q4圖解說明源極_沒極電流1〇與控制閑極電壓 vCG之間的關係。這四個實心lDfiVeG曲線代表可程式化至 一記憶體單元之一淬飭閽搞μ + ,于動閘極上之四個可能的電荷位準,其 分別對應於四個可能的記憶體狀態。作為H —單元 群體之一臨限電屋窗可…至3.5 V範圍内。可藉由以約 〇·4 V母個之間隔將該臨限窗劃分成八個區來分界分別代 變一個經擦除狀態及七個經程式化之狀態之七個可能的記 憶體狀態,,2"、"3,,、,,4"、”5”、%”及”7,、舉例 而言、,若如所圖示使用一為〇.〇5uA之參考電流贿,則 可將以Q1程式化之單元視為處於記憶體狀態"1"中,此乃 因其曲線在由VCG=0.43 ¥及〇88 v分界之臨限窗區之" 與IREF相交。類似地,Q4處於一記憶體狀態"5”中。 由以上描述可看出,使一記憶體單元儲存之狀態愈多, ^限值㈣分得愈精細。舉例而言,_記憶體器件可具 :具有-在-1·5 V至5 V範圍内之臨限窗之記憶體單元。此 提供為6·5 V之最大寬度。若該記憶體單元將儲存16個狀 13I985.doc 15 200921679 態,則每一狀態皆可在該臨限窗中佔據自35〇…至彻mV。 此將需要更高的程式化及讀取操作精確度,以便能夠達成 所要求之解析度。 圖4圖解說明一N0R記憶體單元陣列之一實例。在記憶 體陣列200中’每-列記憶體單元由其源極} 4及沒極! 6以 —菊鏈方式連接。此設計#時稱為一虛接接地設計。一列 中之單7L 10使其控制閘極3〇連接至一字線(例如,字線 42) 一行中之單元使其源極及汲極分別連接至選定位元 線(例如,位元線34及3 6)。 圖5 A示意性地圖解說明組織成一 NAND串之一串記憶體 單元 NAND串50由一系列藉由其源極及汲極菊鏈之記 憶體電晶體Ml,M2,…Μη(例如,n=4, 8, 16或更高)組成。 一對選擇電晶體SI、S2分別經由NAND串之源極端子54及 汲極端子56控制記憶體電晶體鏈與外部之連接。在一記憶 體陣列中,在源極選擇電晶體S1接通時,該源極端子耦合 至一源極線(見圖5B)。類似地,在汲極選擇電晶體S2接通 時’該NAND串之汲極端子耦合至記憶體陣列之一位元 線該鏈中之母一記憶體電晶體1 0皆充當一記憶體單元。 其具有一電荷儲存元件20來儲存一既定量之電荷量以代表 一期望之記憶體狀態。每一記憶體電晶體之—控制閘極3〇 白允許對讀取及寫入操作進行控制。如將在圖5B中所見, 一列NAND串之對應記憶體電晶體之控制閘極3〇所有皆連 接至相同字線。類似地’選擇電晶體S 1、S2中之每一者之 一控制閘極32分別經由其源極端子54及汲極端子%控制控 131985.doc 16 200921679 制對該NAND串之存取。同樣地,一列NAND串之對應選 擇電晶體之控制閘極32所有皆連接至相同選擇線。 、 在在程式化期間讀取或驗證一 NAND串内之—經定址之 δ己憶體電晶體1〇時,給其控制閘極3〇供應一適當之電壓。 同時,藉由在其控制閘極上施加足夠的電壓完全接通 NAND串50中之剩餘未經定址之記憶體電晶體。以此方 式,有效地自個別記憶體電晶體之源極向該nand串之源 極端子54創建-導電路徑,且同樣地自個別記憶體電晶體 之汲極向該單元之汲極端子56創建一導電路徑。美國專利 第 5,570,3 15、5,903,495 及 6,046,935 號中描述了 具有此 NAND串結構之記憶體器件。 圖5B圖解說明一由例如圖5 a中所示之nand串5〇構成之 NAND記憶體單元陣列2⑼之—實例。沿每—行财助串, 一位元線(例如,位元線36)耦合至每一 NAND串之汲極端 子56。/0每一排NAND串,一源極線(例如,源極線μ)耦 合至每一 NAND串之源極端子54。同樣,沿一排^_串 中之-列記憶體單元之控制閘極皆連接至一字線(例如, 字線42)。沿-排NAND串中之一列選擇電晶體之控制閑極 皆連接至一選擇線(例如,選擇線44)。一排nand串中之 一整列記憶體單元可藉助該排NAND串之字線及選擇線上 之適當之電壓來定址。在正讀取-nand串内之一記憶體 電晶體時,該串中之剩餘記憶體電晶體經由其相關聯之字 線強接通’以使流過該串之電流實質上相依於儲存於正讀 取之單元中之電荷位準。 131985.doc ,17- 200921679 程式化及驗證 圖6圖解°兒明用於藉助一系列交替程式化/驗證循環將— °己隐體單兀頁程式化至-目標記憶體狀態之典型技術。經 由耦σ之予線,將一程式化電壓vPGM施加至記憶體單元 之控制閘極。该VPGM係一系列以一自一初始電壓位準 開始之樓梯波形形式之程式化電壓脈衝。正程式化 之早兀經受此系列程式化電壓脈衝,同時每次試圖向浮動 m 加曰加之電荷。在程式化脈衝之間,讀回或驗證該 單元,以相對於一斷點位準確定其源極-汲極電流。該讀 回過程可涉及一個或多個感測操作。在已驗證該單元達到 目‘狀恕時’其程式化停止。所使用之程式化脈衝序列可 具有不斷增加之週期或振幅,以抵消程式化至該記憶體單 元之電荷儲存單元中之累積電子。一般而言,程式化電路 向一選定字線施加一系列程式化脈衝。以此方式’可一起 程式化其控制閑極搞合至字線之一記憶體單元頁。只要已 U ㈣頁之—記憶體單元程式化至其目標狀態,即禁止其程 式 ❿其他單70繼續經受帛式化直JL已驗證該頁之所有 單元皆經程式化。 有 記憶體狀態劃分之實例 ”圖7⑴圖解說明—具有_經擦除狀態作為—接地狀態 Gr及漸進地更多經程式化記憶體狀態,,A"、,,B"及”C"之 實例性4狀態記憶體陣列之臨限電壓分佈。在讀取期間, 藉助三個分界斷點0為分界四個狀態。 θ (2)圖解說明較佳、2位元講編碼來代表圖7⑴中 131985.doc -18· 200921679 所示之該四個可At从 了此的记憶體狀態。分別用一對”上部, 部碼位元(亦即,"1 1 ” " π 1 " 11 、 01 、,'〇〇"及"10")來代表該等記 憶體狀(亦即,"Γ ”1 illustrates a preferred arrangement in which a plurality of peripheral electrical material rows are accessed on opposite sides of the memory array 200 in a symmetrical manner to halve the density of the access lines and circuits of each side. . Because &, the column decoder is divided into column decoders 2 and 2, and the row decoder is divided into row decoders 260A and thin. In the actual case where the column memory cell is divided into multiple pages'! The rural worker 250 is divided into page multiplexers 25, 8 and 2 times. Similarly, the read/write circuit 270 is divided into a read/write circuit 270A connected from the bottom to the bit line and a read/write circuit 270B connected from the top to the bit line. In this manner, the density of the read/write modules, and thus the density of the sense module 380, is substantially halved. Control circuit 110 is an on-wafer controller that cooperates with read/write circuit 270 to perform memory operations on memory array 200. Control circuit 110 typically includes a state machine 112 and other circuitry (e.g., an on-wafer address decoder and a power control module (not explicitly shown)). The state machine ιΐ2 provides wafer level control for memory operations. The control circuit communicates with a host via an external memory controller. Memory array 200 is typically organized as a two-dimensional array of memory cells arranged in columns and rows and addressable by word lines and bit lines. The array can be formed according to a 131985.doc 13 200921679 nor type or a NANr^M type architecture. Figure 2 is an illustration of a non-volatile memory unit. The memory cell 10 can be constructed from a field effect transistor having a charge storage unit 2 (e.g., a floating gate or a dielectric layer). The memory unit 1A also includes a source 14, a pole 16 and a control gate 3〇. Many commercially available non-volatile solid state memory devices are being used today. The memory devices can employ different types of memory cells, each having one or more charge storage elements. Typical non-volatile memory cells include eepr〇m and flash EEPR0M. An example of an EEpR〇M unit and a method of manufacturing the same are given in U.S. Patent No. 5,595,924. Examples of flash EEPROM cells, their use in memory systems, and methods of making the same are given in U.S. Patent Nos. 5, 〇 7,32, 5,95,344, 5,315,541, 5,343,063, 5,661,053, 5,313,421, and 6,222,762. . In particular, examples of memory devices having a unit structure are described in U.S. Patent Nos. 5,57,315, 5,903,495 and 6,046,935. Similarly, Eitan et al. have been "NR〇M: A Novel Localized Trapping, 2-Bit Nonvolatile Memory Cell, " (IEEE Electron Device Letters, Vol. 21, No. U, (9) (Eight) Year ^ Month, No. 543-S45 An example of a memory device utilizing a dielectric storage element is described in U.S. Patent No. 5,768,192 and U.S. Patent No. 6,725. In practice, the memory state of the cell is typically read by sensing the conduction current across a single 7L source and drain electrode when a reference voltage is applied to the control gate. Therefore, for each of the predetermined charges on the floating gate of a cell, a corresponding voltage can be detected relative to a fixed reference control. The voltage is 131985.doc •14· 200921679 = current. Similarly, the range of charge that can be programmed to the floating idler corresponds to the S-limit voltage window or a corresponding conduction current window. Another-selection, no--detects the conducted electricity in the divided current window, but can set the threshold current for a given memory state at the control closed-end: and detect whether the conduction current is lower or higher A limit current. In one embodiment, the conduction current is detected relative to the - threshold current by checking the rate at which the conduction current is discharged through the capacitance of the bit line. Ο u Figure Addition The four different outputs Q1-Q4 that can be selectively stored at any time for the floating gate illustrate the relationship between source _ immersion current 1 〇 and control idle voltage vCG. These four solid lDfiVeG curves represent four possible charge levels that can be programmed into one of the memory cells, which correspond to the four possible memory states, respectively. As one of the H-unit groups, the electricity house window can be ... to 3.5 V. The seven possible memory states of an erased state and seven stylized states can be respectively demarcated by dividing the threshold window into eight regions at intervals of about 〇·4 V. , 2", "3,,,,,4","5",%", and 7,7, for example, if the use of one is 〇.〇5uA reference current bribe, as shown The unit stylized in Q1 is considered to be in the memory state "1" because its curve intersects IREF in the threshold window of the boundary between VCG=0.43 ¥ and 〇88 v. Similarly, Q4 is in a memory state "5" As can be seen from the above description, the more states a memory cell is stored, the finer the limit value (4) is. For example, the memory device It can be: a memory unit with a threshold window in the range -1·5 V to 5 V. This provides a maximum width of 6.5 V. If the memory unit will store 16 shapes, 13I985.doc 15 In the 200921679 state, each state can occupy from 35〇...to mV in the threshold window. This will require higher stylization and read operation accuracy to achieve the required resolution. An example of an array of NOR memory cells is illustrated. In the memory array 200, 'per-column memory cells are connected by their sources} 4 and no poles! 6 is connected in a daisy-chain manner. This design is called a The virtual grounding design. The single 7L 10 in one column has its control gate 3〇 connected to a word line (for example, word line 42). The cell in one row connects its source and drain to the selected positioning element (eg , bit lines 34 and 3 6). Figure 5A schematically illustrates a string of memory organized into a NAND string The unit NAND string 50 is composed of a series of memory transistors M1, M2, ... Μη (for example, n = 4, 8, 16 or higher) which are daisy-chained by their sources and spurs. A pair of selective transistors SI S2 controls the connection of the memory transistor chain to the outside via the source terminal 54 and the NMOS terminal 56 of the NAND string. In a memory array, the source terminal is coupled when the source selective transistor S1 is turned on. To a source line (see Figure 5B). Similarly, when the drain select transistor S2 is turned on, the 汲 terminal of the NAND string is coupled to one of the bit lines of the memory array. Each of the transistors 10 functions as a memory cell. It has a charge storage element 20 for storing a predetermined amount of charge to represent a desired memory state. Each memory transistor - control gate 3 is allowed to be white The read and write operations are controlled. As will be seen in Figure 5B, the control gates 3 of the corresponding memory transistors of a column of NAND strings are all connected to the same word line. Similarly 'select transistor S1, One of each of S2 controls gate 32 via its source terminal 54 and its drain Terminal % Control 131985.doc 16 200921679 The access to the NAND string is made. Similarly, the control gates 32 of the corresponding select transistors of a column of NAND strings are all connected to the same select line. Or verifying that the addressed δ 忆 体 电 电 , , , , , , , , , , , , , , , , 验证 验证 验证 验证 验证 验证 验证 验证 验证 验证 验证 验证 验证 验证 验证 验证 验证 验证 验证 验证 验证 验证 验证 验证 验证 验证 验证The remaining unaddressed memory transistors in the NAND string 50. In this manner, a conductive path is effectively created from the source of the individual memory transistors to the source terminals 54 of the nand string, and likewise from individual memories The drain of the bulk transistor creates a conductive path to the germanium terminal 56 of the cell. A memory device having such a NAND string structure is described in U.S. Patent Nos. 5,570,315, 5,903,495 and 6,046,935. Figure 5B illustrates an example of a NAND memory cell array 2 (9) constructed of, for example, the nand string 5 图 shown in Figure 5a. Along the per-banking string, a bit line (e.g., bit line 36) is coupled to the 汲 terminal 56 of each NAND string. /0 Each row of NAND strings, a source line (e.g., source line μ) is coupled to source terminal 54 of each NAND string. Similarly, the control gates of the column memory cells along a row of strings are connected to a word line (e.g., word line 42). The control idlers of the selected transistor along one of the row-row NAND strings are all connected to a select line (e.g., select line 44). An entire column of memory cells in a row of nand strings can be addressed by the word lines of the row of NAND strings and the appropriate voltage on the select lines. While a memory transistor in the -nand string is being read, the remaining memory transistors in the string are strongly "on" via their associated word lines such that the current flowing through the string is substantially dependent on the stored The level of charge in the cell being read. 131985.doc , 17- 200921679 Stylization and Verification Figure 6 illustrates a typical technique for stylizing a single-page of a hidden body into a target memory state by means of a series of alternating stylization/verification cycles. A stylized voltage vPGM is applied to the control gate of the memory cell via a pre-coupled σ. The VPGM is a series of stylized voltage pulses in the form of a staircase waveform starting from an initial voltage level. The stylized early is subject to this series of stylized voltage pulses, while each attempting to add a charge to the floating m. Between the stylized pulses, the cell is read back or verified to determine its source-drain current relative to a breakpoint level. The readback process can involve one or more sensing operations. Stylized stop when the unit has been verified to have reached the goal. The stylized pulse sequence used can have an increasing period or amplitude to counteract the accumulated electrons that are programmed into the charge storage unit of the memory unit. In general, the stylized circuit applies a series of stylized pulses to a selected word line. In this way, the memory cells can be programmed together to control one of the memory cells. As long as the U (four) page - memory unit is programmed to its target state, its program is disabled. Other singles 70 continue to undergo normalization. JL has verified that all units of the page have been programmed. An example of memory state partitioning is illustrated in Figure 7(1)—with the _erased state as the grounded state Gr and progressively more stylized memory states, A",,,B" and "C" The threshold voltage distribution of the 4 state memory array. During the reading, four states are delimited by three boundary breakpoints 0. θ (2) illustrates a preferred, 2-bit speaker code to represent the memory states of the four At which are shown in Figure 13(1), 131985.doc -18·200921679. Use a pair of "upper, partial code bits (ie, "1 1 " " π 1 " 11 , 01 , , '〇〇" and "10") to represent the memory ( That is, "Γ ”

Gf、"Α"、"Β"及”c”)中之每— 美國專利第6,657 891 % 士 # ,91唬中揭示"LM"碼,且其因避免需要 大的電荷改變的藉4& 式^操作而有利於減小此鄰浮動閘極之 間之場效輕合。將兮短zt洲· 將忒編碼玟计成可分開程式化及讀取2個 :位兀(下。|5 &上部”位元)。在程式化下部位元時,該Each of Gf, "Α", "Β" and "c") - US Patent No. 6,657 891 % 士#, 91 揭示 reveals "LM" code, and it avoids the need for large charge changes The 4&^ operation facilitates reducing the field effect of the adjacent floating gates. It will be shortened ztzhou. The 忒 code is counted as separable and can be read separately: position 兀 (below. |5 & upper bit). When stylized part of the element,

單元之臨限位準維持在"經擦除”區中或向臨限窗之一"中 下區移動。在程式化上部位元時,該兩個區中一 t 中之-單元之臨限位準進一步提高至該臨限窗之一”下中 部”區中之一稍微較高位準。 圖8⑴圖解說明-實例性8狀態記憶體陣列之臨限電壓 分佈。每-記憶體單元之可能的臨限電壓跨越一被劃分成 八個區之臨限由,以分界八個可能的記憶體狀態"Ο〆,、 "Α" ' "Β" ' π ' "D" ' "E” ' nF"及"G”。”&"係一接地狀 態,其係一緊密式分佈内之一經擦除狀態,且"八”-”…係 七個漸進地經程式化之狀態。在讀取期間,藉助七個分界 斷點DA-DG分界該八個狀態。 圖8(2)圖解說明-較佳、3位元編碼來代表圖8(丨)中所示 之該八個可能的記憶體狀態。分別用一三位元組”上部、 中間、下部”(亦即,"m"、”011"、"_,,、、 ""、"_··、’實’及"m")來代表該人個記憶體狀態中 之每一者。w言亥編碼設計成可分開料化及讀取3個碼位 元”下部,,、”中間”及”上部”位元。因此,第一回合,若下 131985.doc •19- 200921679 部位元為” 1 ”,則T加π h ρ頁程式化使一單元維持在,,經擦除"或 ^ ,占’或若下部位元為"〇”,則程式化至中下部"狀 ^。土本上’,,Gr”或”接地”狀態因具有程式化至一狹窄臨 限值範圍內夕M、柯& . 乍 _ 、,,冰度擦除狀態而係具有一緊密式分佈之 系工擦除'狀態。"中下Λη^呈古 下σ卩狀態可具有一検跨於記憶體狀態 之間之廣闊臨限電壓分佈。在程式化期間,可相 對於一粗略斷點臨限位準(例如,db)驗證,,中下部"狀態。 在私式化中間位元時,一單元之臨限位準將自由下部頁程 式化產生之兩個區中之一者開始並向四個可能的區中之一 者移動。在程式化上部位元時,一單元之臨限位準將自由 中間頁程式化產生之四個可能的區中之—者開始並向八個 可能的記憶體狀態移動。 感測電路及技術 圖9圖解說明圖丨中所示之跨越一記憶體單元陣列含有一 排p個感測模組之讀取/寫入電路270八及27〇B。整排並行地 操作之p個感測模組480允許將並行地讀取或程式化一沿一 行之P個單元10之塊(頁)。實質上,感測模組丨將感測單元丄 中之一電流I,,感測模組2將感測單元2中之一電流 12,...,感測模組P將感測單元p中之一電流Ip等等。該頁 之流出源極線34到達一聚集節點CLSRC中並自此到達接地 之總單元電流將係該p個單元中之所有電流之一總和。 在駕用S己憶體架構中,一列具有一共用字線之記憶體單元 形成兩個或更多個頁’其中並行地讀取及程式化一頁中之 δ己憶體單元。在一具有兩個頁之列之情形下,由偶數位元 131985.doc -20- 200921679 線存取一個頁,且由奇數位元線存取其他頁。一感測電路 頁在任一時刻耦合至偶數位元線或奇數位元線。 在當前生產之基於56 nm技術之晶片中p > 64000且在43 nm 32 Gbit x4晶片中p > 150000。在較佳實施例中,該塊 為一連串整列記憶體單元。此係所謂的”所有位元線”架 構,其中該頁係由一列分別耦合至相鄰位元線之相鄰記憶 體單元構成。在另一實施例中,該塊為該列中之單元之一 子組。舉例而言,該單元子組可為整列的一半或整列的四 分之一。該單元子組可為一連串相鄰單元或每隔另一單元 一個,或每個一預定數目單元一個的串。每一感測模組經 由一位元線耦合至一記憶體單元且包含一用於感測一記憶 體單元之傳導電流之感測放大器。一般而言,若讀取/寫 入電路分佈於該記憶體陣列之相對側上,則該排p個感測 模組將分佈於兩組讀取/寫入電路270A與270B之間。已在 Cernea等人於2005年8月4曰申請之題為"IMPROVED MEMORY SENSING CIRCUIT AND METHOD FOR LOW VOLTAGE OPERATION”之美國專利公開案第2005-0169082-A1號中揭示較佳感測模組。美國專利公開案第 2005-01 69082-A1號之全部揭示内容以引用方式併入本文 中〇 圖1 0更詳細地示意性地圖解說明圖9中所示之適宜於實 踐本發明之感測模組。感測模組4 8 0經由一所耗合之位元 線36感測一 NAND鏈50中之一記憶體單元之傳導電流。其 具有一可選擇性地耦合至一位元線之感測節點48 1、一感 131985.doc -21 - 200921679 測放大器600及一讀出匯流排499。初始,一隔離電晶體 482在藉由一信號8^啓用時將位元線“連接至感測節點 481。感測放大器600對感測節點481進行感測。該感測放 大器包含一預充電/鉗位電路64〇、一單元電流鑑別器65〇 及一鎖存器660。感測模組480使得能夠感測該NAND鏈中 之選定記憶體單元之傳導電流。 在一較佳實施例中,提供一下拉電路55〇以選擇性地將 位元線36拉至接地。下拉電路55〇在信號INv及另一信號 GRS兩者皆為高時啓動。供應由狀態機丨丨〇(見圖丨)控制之 b號GRS作為來自頁控制器498之控制及定時信號之一部 分。如稍後將更詳細描述’信號GRS可由狀態機視為控制 b號來啓用(GRS=高)或停用(GRS =低)下拉電路550以分別 啓用或停用高電流位元線之鎖定。在該感測指示一高電流 狀態時’ INV將為高,且若啓用該下拉電路時,其將下拉 位元線。 在感測之前’必須在一個或多個預充電操作中經由適當 之字線及位元線來設定到達選定記憶體單元之閘極之電 壓。舉例而言’如圖10中所示,可選擇一沿一與NAND鏈 50相交之字線WL1之記憶體單元頁以進行感測。預充電操 作自正充電至一電壓Vread之未選擇之字線WL0、WL2_ WL3 1及正針對一考慮中之既定記憶體狀態充電至一預定 臨限電壓VT(i)之選定字線WL1開始。接著,位元線預充電 電路6 4 0使位元線3 6到達一適於感測之預定沒極電麼。此 將促使一源極—汲極傳導電流在NAND鏈50中之選定記憶 I31985.doc 22· 200921679 體單元中流動,該源極-汲極傳導電流係經由一所耦合之 位元線36從該NAND鏈之通道偵測到。該傳導電流係在在 記憶體單元之源極與汲極之間存在一標稱電壓差時程式化 至選定記憶體單元中之電荷及所施加之VT(i)之一函數。 圖11A更詳細地圖解說明圖1〇所示之預充電/钳位電路。 该電路具有一電壓鉗位620'組件及一預充電電路640,組 件。電壓鉗位620’係由一由其閘極處之一類比信號Β[χ加 以控制之電晶體612構建。BLX在節點SEN2 481(見圖1〇)上 確保足夠的電壓以使位元線電壓鉗位6丨〇可適當地運行。 在VT(i)電壓穩定時,可經由所耦合之位元線36經由由一 b號XXL閘控之電晶體63 〇來感測選定記憶體單元之傳導 電流或經程式化之臨限電壓。單元電流鑑別器65〇充當電 流位準之一鑑別器或比較器。其耦合至感測節點以感測該 記憶體單元中之傳導電流。 圖11B更詳細地圖解說明圖1〇中所示之單元電流鑑別器 電路。單70電流鑑別器65〇,包含一電容器652及一 p通道電 晶體656。該單元電流鑑別器實質上根據其使電容器652充 電或放電之速率來量測一記憶體記憶體單元之傳導電流。 此藉由在節點631處感測信號SEN來完成。信號SEN控制p 電體656之閘極。在感測之前,藉助預充電電路6糾,將 SEN預充電至vdd(高)。亦參照圖10,藉助—接通耦合電晶 體632以使節點SEN 651在節點647處耦合至預充電電路之 信號HHL來啓用預充電。此將將跨越電容器652之電壓初 始設定至零。接著藉由根據該單元使該電容器放電之速率 131985.doc •23 - 200921679 來量測其之傳導電流完成感測。 在感測期間’位元線中之記憶體單元之傳導電流將使電 容器652放電。節點SE种之電塵將接著以-相依於傳導 電流之速率自vdd減少。在一預 頂疋放電週期之後(該週期對 應於一參考電流),SEN將降低至某—可接通或可不接通量 •電晶體656之值。若其降低至足以接通P電晶體656,則 此將意味著傳導電流高於參考電流。此亦將導致⑽ 在斷定信號STB時拉高。另-方面’若在感測週期結;時 未接通電晶體656,則傳導電流低於參考電流且信號請將The threshold level of the unit is maintained in the "erased" area or moved to one of the threshold windows "lower and middle areas. In the stylized upper part, one of the two areas The threshold level is further increased to a slightly higher level in one of the "lower middle" regions of the threshold window. Figure 8 (1) illustrates the threshold voltage distribution of the exemplary 8-state memory array. The possible threshold voltage spans a threshold divided into eight zones to demarcate eight possible memory states "Ο〆,, "Α" ' "Β" ' π ' "D" ' "E" 'nF" and "G"."&" is a grounded state, which is an erased state in a tightly distributed state, and "eight"-"... is seven progressively Stylized state. During reading, the eight states are delimited by seven boundary breakpoints DA-DG. Figure 8 (2) illustrates - preferred, 3-bit encoding to represent the eight possible memory states shown in Figure 8 (丨). Use a three-tuple "upper, middle, lower" (ie, "m","011", "_,,,, "", "_··, 'real' and &quot ;m") to represent each of the memory states of the person. w hai code is designed to be separate and read 3 code bits "lower,", "middle" and "upper" bits . Therefore, in the first round, if the 131.3.doc •19- 200921679 part element is “1”, then T plus π h ρ page stylizes to maintain a unit, erased " or ^, accounted for 'or The lower part is "quote", which is stylized to the middle and lower part. The ",", "Gr" or "grounded" state is studded to a narrow threshold. M, Ke &; 乍 _ , , ,, ice-erased state with a close-distributed system erase state. "The lower middle Λ 呈 ^ is the ancient σ 卩 state can have a wide threshold voltage distribution across the state of the memory. During stylization, it can be verified against a coarse breakpoint threshold level (for example, db), the lower middle " state. When the intermediate bit is privately categorized, the threshold of one unit begins with one of the two regions resulting from the lower page programming and moves to one of the four possible regions. When stylizing the upper part, a unit's threshold level starts with the four possible areas of the free intermediate page stylization and moves to eight possible memory states. SENSE CIRCUIT AND TECHNIQUE Figure 9 illustrates the read/write circuits 270 and 27B of a memory cell array having a row of p sensing modules as shown in the Figure. The entire row of p sensing modules 480 operating in parallel allows parallel reading or programming of blocks (pages) of P cells 10 along a row. In essence, the sensing module 丨 will sense one of the currents I in the unit ,, the sensing module 2 will sense one of the currents 12 in the sensing unit 2, ..., the sensing module P will sense the unit p One of the currents Ip and so on. The total cell current flowing from the source line 34 of the page to a gathering node CLSRC and from there to ground will be the sum of all of the currents in the p cells. In a driving S-memory architecture, a column of memory cells having a common word line forms two or more pages' in which the δ-resonant cells in a page are read and programmed in parallel. In the case of a page with two pages, one page is accessed by the even bit 131985.doc -20-200921679 line, and the other pages are accessed by the odd bit line. A sense circuit page is coupled to an even bit line or an odd bit line at any one time. In the currently produced 56 nm technology based wafer p > 64000 and in the 43 nm 32 Gbit x 4 wafer p > 150000. In the preferred embodiment, the block is a series of aligned memory cells. This is a so-called "all bit line" architecture in which the page consists of a column of adjacent memory cells respectively coupled to adjacent bit lines. In another embodiment, the block is a subset of the cells in the column. For example, the unit sub-group can be half of the entire column or a quarter of the entire column. The unit sub-group may be a series of adjacent units or one string every other unit, or one string each of a predetermined number of units. Each sensing module is coupled to a memory cell via a bit line and includes a sense amplifier for sensing the conduction current of a memory cell. In general, if the read/write circuits are distributed on opposite sides of the memory array, the rows of p sense modules will be distributed between the two sets of read/write circuits 270A and 270B. A preferred sensing module is disclosed in U.S. Patent Publication No. 2005-0169082-A1, the entire disclosure of which is incorporated herein by reference. The entire disclosure of U.S. Patent Publication No. 2005-01 69082-A1 is incorporated herein by reference in its entirety herein in its entirety in in in in in in in in in in in The sensing module 480 senses a conduction current of a memory cell in a NAND chain 50 via a consuming bit line 36. It has a sense of being selectively coupled to a bit line Measure node 48 1 , a sense 131985.doc -21 - 200921679 amp 600 and a read bus 499. Initially, an isolation transistor 482 "connects" the bit line to the sense when enabled by a signal 8^ Node 481. The sense amplifier 600 senses the sense node 481. The sense amplifier includes a precharge/clamp circuit 64A, a cell current discriminator 65A, and a latch 660. Sensing module 480 enables sensing of the conduction current of selected memory cells in the NAND chain. In a preferred embodiment, a pull-down circuit 55 is provided to selectively pull the bit line 36 to ground. The pull-down circuit 55 is activated when both the signal INv and the other signal GRS are high. The b-number GRS controlled by the state machine (see Fig. 供应) is supplied as part of the control and timing signals from the page controller 498. As will be described in more detail later, the 'signal GRS can be enabled by the state machine as a control b number to enable (GRS = high) or disable (GRS = low) pull down circuit 550 to enable or disable the locking of the high current bit line, respectively. When the sense indicates a high current state, 'INV will be high, and if the pull-down circuit is enabled, it will pull down the bit line. Prior to sensing, the voltage to the gate of the selected memory cell must be set via the appropriate word line and bit line in one or more pre-charge operations. For example, as shown in FIG. 10, a memory cell page along a word line WL1 intersecting the NAND chain 50 can be selected for sensing. The precharge operation begins with the unselected word lines WL0, WL2_WL3 1 being positively charged to a voltage Vread and being selected for a predetermined memory state to be charged to a predetermined threshold voltage VT(i). Next, the bit line precharge circuit 640 causes the bit line 36 to reach a predetermined non-polarity suitable for sensing. This will cause a source-drain conduction current to flow in the selected memory of the NAND chain 50, the source-drain conduction current being passed through a coupled bit line 36. The channel of the NAND chain is detected. The conduction current is programmed as a function of the charge in the selected memory cell and the applied VT(i) when there is a nominal voltage difference between the source and drain of the memory cell. Figure 11A illustrates the precharge/clamp circuit shown in Figure 1A in more detail. The circuit has a voltage clamp 620' component and a precharge circuit 640, component. The voltage clamp 620' is constructed from an analog signal 612 [χ controlled by a transistor 612 at its gate. BLX ensures sufficient voltage on node SEN2 481 (see Figure 1A) to clamp the bit line voltage to 6 丨〇 to operate properly. When the VT(i) voltage is stable, the conduction current or the programmed threshold voltage of the selected memory cell can be sensed via the coupled bit line 36 via the transistor 63 闸 gated by a b-th XXL. The cell current discriminator 65A acts as a current discriminator or comparator. It is coupled to the sensing node to sense the conduction current in the memory unit. Figure 11B illustrates the cell current discriminator circuit shown in Figure 1A in more detail. The single 70 current discriminator 65A includes a capacitor 652 and a p-channel transistor 656. The cell current discriminator measures the conduction current of a memory memory cell substantially at a rate at which it causes the capacitor 652 to charge or discharge. This is done by sensing the signal SEN at node 631. Signal SEN controls the gate of p-electrode 656. Prior to sensing, SEN is precharged to vdd (high) by means of precharge circuit 6. Referring also to Figure 10, pre-charging is enabled by turning on the coupled transistor 632 to cause the node SEN 651 to couple to the pre-charge circuit signal HHL at node 647. This will initially set the voltage across capacitor 652 to zero. The sensing is then measured by measuring the conduction current of the capacitor according to the rate at which the capacitor is discharged according to the unit 131985.doc • 23 - 200921679. The conduction current of the memory cells in the bit line during sensing will cause capacitor 652 to discharge. The electric dust of the node SE will then decrease from vdd at a rate dependent on the conduction current. After a pre-top 疋 discharge period (which corresponds to a reference current), SEN will decrease to a value that can be turned on or off • the value of transistor 656. If it is reduced enough to turn on P transistor 656, this would mean that the conduction current is higher than the reference current. This will also cause (10) to pull high when the signal STB is asserted. Another aspect: if the transistor 656 is not turned on during the sensing period; the conduction current is lower than the reference current and the signal is

為低。亦參照圖10,藉由藉助斷開耦合電晶體63〇之XXL 自SEN節點退耗位元線來標記感測週期之結束。接著藉助 一選通信號STB將所感測之結果鎖存至鎖存中。 單元電流鑑別器650有效地確定單元之傳導電流是高於 或低於既定分界電流值。該既定分界電流值對應於一預 疋放電時間。右所感測之電流高於該分界電流值,則將鎖 存“60設定至一預定狀態’其十信號胸=1(高)。此亦意 未著讨_中之s己憶體單元在控制閘極處具有一小於所施加 之VT(i)之臨限值。 一般而言,將存在一正由一對應數目之多通過感測模組 480處理之記憶體單元頁。一頁控制器a%向該等感測模組 中之每一者供應控制及定時信號。頁控制器498使多通過 感測杈組480中之每一者循環通過—預定數目之通過〇 = 1至 N)且亦為每一個通過供應一預定分界電流值〖ο⑴。如在此 技術中所衆所周知,亦可將分界電流值實施為一分界臨限 131985.doc -24- 200921679 電壓或用於感測之時間週期。在最後一個通過之後,頁控 制器498藉助一信號NCO啓用一傳輸閘極488,以將感測節 點481之狀態作為所感測之資料讀取至讀出匯流排499。總 之’將自所有多通過模組48〇讀出一感測資料頁。 感測期間之高電流記憶體單元之問題 如先前所述,為増加讀取效能,並行地感測一記憶體單 疋頁,且該頁越大效能越高。然而,如圖9顯而易見,並 行地操作大量單元亦將消耗大量電流。 許多問題由以大量電流操作引起。一般而言,始終期望 具有一消耗較少功率之器件。特定而言,必須容納較高電 流之組件將很可能更為笨重且佔據有價值的晶片空間。常 常,s己憶體器件係針對更差情形電流而設計,而大多數時 間正操作甚少之電流。此乃因電流係相依於程式化至該等 單元中之資料,較少經程式化之單元具有較高傳導電流。 另一問題涉及由源極線與晶片之接地銲墊之間之一有限 電阻引入之一誤差。感測記憶體單元之一個電位問題係由 跨越該有效電阻之源極負载引起之源極線偏壓。在並行地 感測大量§己憶體單元時,其經組合之電流可在一具有有限 電阻之接地迴路中導致顯著的電壓降。此導致一源極線偏 壓,該源極線偏壓在一採用臨限電壓感測之讀取操作中引 起誤差。 圖12A圖解說明由於具有—有限對地電阻之源極線中之 電流而引起之源極電壓誤差問題。讀取/寫入電路而及 2观同時處理一記憶體單元頁。該等讀取/寫入電路中之 131985,doc -25- 200921679It is low. Referring also to Figure 10, the end of the sensing period is marked by de-energizing the bit line from the SEN node by disengaging the XXL of the coupling transistor 63A. The sensed result is then latched into the latch by means of a strobe signal STB. The cell current discriminator 650 effectively determines whether the conduction current of the cell is above or below a predetermined demarcation current value. The predetermined demarcation current value corresponds to a pre-emission discharge time. If the current sensed by the right is higher than the value of the demarcation current, the latch "60 is set to a predetermined state" and its ten signal chest = 1 (high). This is also intended to be in the control. The gate has a threshold less than the applied VT(i). In general, there will be a page of memory cells being processed by the sensing module 480 by a corresponding number. % supplying control and timing signals to each of the sensing modules. The page controller 498 cycles each of the multiple pass sensing sets 480 - a predetermined number of passes 〇 = 1 to N) and A predetermined demarcation current value is also supplied for each pass (1). As is well known in the art, the demarcation current value can also be implemented as a demarcation limit of 131985.doc -24-200921679 voltage or for sensing Time period. After the last pass, page controller 498 enables a transmit gate 488 by means of a signal NCO to read the state of sense node 481 as sensed data to read bus 499. In short, 'will All multi-pass module 48〇 reads a sensing data page. The problem of the high current memory unit is as described above, in order to increase the read performance, a memory single page is sensed in parallel, and the page is more powerful. However, as is apparent from Fig. 9, a large number of operations are performed in parallel. The unit will also consume a large amount of current. Many problems are caused by operation with a large amount of current. In general, it is always desirable to have a device that consumes less power. In particular, components that must accommodate higher currents will likely be more cumbersome and occupy Valuable wafer space. Often, suffix devices are designed for worse currents, and most of the time is operating with less current. This is because the current system is dependent on the data programmed into the cells. A less-programmed unit has a higher conduction current. Another problem involves introducing one of the finite resistances between the source line and the ground pad of the wafer. One potential problem of the sense memory cell is crossed. The source line bias caused by the source load of the effective resistor. When a large number of § memory cells are sensed in parallel, the combined current can have a finite resistance A significant voltage drop is caused in the ground loop. This results in a source line bias that causes an error in a read operation with threshold voltage sensing. Figure 12A illustrates the presence of a limited-to-ground The source voltage error caused by the current in the source line of the resistor. The read/write circuit and the 2 memory simultaneously process a memory cell page. 131985, doc -25 in the read/write circuit - 200921679

每一感測模組480皆經由一位元線36耦合至—對應單元。 舉例而言,一感測模組480感測一記憶體單元1〇之傳導電 流Μ源極-汲極電流)。該傳導電流自該感測模組流過位元 ^36進人記憶體單元1G德極巾,並在通過—源極線^之 月'J自源極14流出到達接地。在一積體電路晶片中,一記悚 體陣列中之單元之祕所有皆連在—起作為源極線34之^、 個支路,源極線34連接至該記憶體晶片之某外部接地銲墊 (例如,Vss銲墊)。即使在使用金屬帶減小源極線之電阻 時,一記憶體單元之源電極與該接地銲墊之間維持一有限 電阻R。通常,接地迴路電阻尺為5〇 〇hm左右。 對於正被並行地感測之整個記憶體頁而言,流過源極線 Μ之總電流係所有傳導電流之總和,亦即,^尸…, lp 般而σ,母一 s己憶體單元皆具有一相依於程式化 至其電荷儲存元件中之電荷量之傳導電流。對於該記憶體 單元之一既定控制閘極電壓而言,少量的電荷將產生一相 對較高的傳導電流(見圖3)。在在一記憶體單元之源電極與 接地銲墊之間存在一有限電阻時,跨越該電阻之電壓降係 由 Vcirop^ Z’rorR給出。 舉例而5,若各自具有一為0.25 μA之電流之24000個位 元線同時放電,則源極線電壓降將等於24〇〇〇個線χ〇 25 μΑ/每條χ50 ohm〜0.3伏。在感測記憶體單元之臨限電壓 時,此源極線偏壓將導致一為0.45伏之感測誤差,假設體 效應使得源極電壓上升0,3 V導致臨限電壓上升〇 45 v。 圖12B圖解說明由一源極線電壓降所引起之記憶體單元 131985.doc -26- 200921679 之臨限電壓位準之誤差。供應至記憶體單元1〇之控制閘極 30之臨限電壓VT係相對於GND。然而,該記憶體單元承受 之有效電壓vT係其控制閘極30與源極14之間之電壓差。在 所供應之ντ與有效%之間存在—約為i 5xvdrcp之差(忽略 自源極14至源極線之較小電壓降影響)。纟感測該等記憶 體早兀之臨限電壓時,此ν“。或源極線偏壓將導致一為 (例如0.45伏)之感測誤差。不可輕易移除此偏壓,乃因其 係資料㈣’亦即相依於該頁之該等記憶體單元之記憶體 狀態。 源極負載及使用位元線鎖定之功率節省技術 已在U等人於2005年3月16日申請之題為1〇义 VOLATILE MEMORY AND METHOD WITH POWER-SAVING READ AND PROGRAM-VERIFY 〇PERATI〇NS”之美國專利申 請案第1 1/083,5 14號中揭示功率節省技術,該專利申請案 之揭示内容以引用方式併入本文中。特定而言,一讀取或 程式化驗證操作包含一個或多個對應於一個或多個分界臨 限電壓之感測循環以確定每一記憶體單元在多個可能的記 憶體狀態中之哪一者中。 每一感測循環係用於相對於一分界臨限電壓進行感測, 並並行地處理一記憶體單元頁。一感測循環通常包含不止 P個通過或子循環來解析該頁中之所有單元之記憶體狀 態。在-個態樣中,一第一個通過或子循環盡可能地感測 並識別頁中具有最高傳導電流之彼等記憶體單元此將最小 化一後續子循環期間由於存在該等高電流單元而引起之任 131985.doc -27- 200921679 7感測誤差。由於已讀取該等單元,故斷開其傳導電流以 即省功率。藉由將其相關聯之位元線接地以使跨越每一單 元之源極及汲極大致不存在電位差來斷開該等單元。在一 後續通過或子循環中,將在來自較高電流單元之減小之干 擾之情形下再次並行地感測該頁之剩餘記憶體單元。 因此,相對於每一分界臨限電壓來區分兩個毗鄰記憶體 狀態,實施至少兩個感測通㉟。第—個通過或子循環係識 別具有低於分界位準之臨限電壓之高電流單元。第二通過 或子循環係在藉由將高電流單元之位元線鎖定至一接地電 位使其斷開之後重複感測。 圖10亦顯示位元線36之一下拉電路55〇 ’下拉電路55〇在 信號INV及另一信號GRS兩者皆為高時啓動。下拉電路55〇 較佳由一第一 η電晶體55〇與一第二n電晶體552串聯構成。 信號INV及GRS在兩者皆為高時將分別接通電晶體55〇及 552。此將將感測節點48 1且亦所連接之位元線36下拉至接 地電壓。不管控制閘極電壓如何,此皆將抑制傳導電流在 記憶體單元1〇中之流動’乃因在其源極與汲極之間將不存 在電壓差。有效地’信號GRS可由狀態機視為控制信號以 啓用(GRS=高)或停用(GRS =低)位元線鎖定。 圖13(Α)·13⑴係—藉助位元線鎖定之2通過感測之定時 圖。應理解,一般而言將存在一用於確定頁中之每一單元 是否具有-低於或高於-分界臨限位準之臨限電壓之感測 循%,該分界臨限位準用於區分兩個記憶體狀態。對於每 —感測循環而言,相對於由該記憶體支援之參考臨限電壓 131985.doc •28· 200921679 中之母一者將存在2個感測通過。 特定°圖i3(A)-13(J)係控制圖1〇中所示之感測模組 彻之操作之信號之定時圖。總體方案係相對於—既定參 考臨限位準或參考傳導電流並行地感測記憶體單元頁。如 先前所述,可藉由相對於一參考電流確定一記憶體單元十 之傳導電流來完成相對於一分界臨限位準感測該單元中之 臨限㈣° —具有—低於該分界臨限位準之臨限電麼之單 元將使其傳導電流高於參考電流。因此,若感測循環以一 遞升次序進行下一分界臨限位準,則每一感測循環皆將區 分具有低於參考傳導電流之傳導電流之彼等單元。 美國專利第7,196,931號揭示一種藉由_2通過感測循環 減小源極偏壓誤差之方法。該2通過感測循環具有識別具 有大致高於參考電流之傳導電流之彼等單元之第一個通 過。在該等單元經識別並斷開之後’在無高電流單元干擾 之情形下在—第二通過中實施相對於大致參考電流 測。 ,Λ 相對於一參考臨限電壓之每一感測因而皆包含分別顯示 為相位(1)-(4)及相位(5)-(9)之至少兩個子循環,其中每— 子循環在並行地感測記憶體單元頁時係—個通過。每—感 測子循環需要一設置以在感測可發生之前使字線及位元線 設定至適當之電壓。此藉由一預充電操作來完成。、、 第一子循環之預充電操作係在相位(丨)_(2)之間且第二 循環之預充電操作係在相位之間。 圖13(A)顯示選定字線之預充電之定時。若感測係相對 131985.doc -29- 200921679 於一為vT1之分界臨限電壓位準,則字線開始預充電至此 電壓位準。端視字線相對於位元線<RC延遲之rc延遲, 子線預充電可開始的比位元線的預充電早。 位元線之預充電可在記憶體單元耦合或不耦合至位元線 之情形下發生。如先前所述,在一個實施例中,該等單元 初始自位元線去耦合以使其汲極電流不防礙位元線之上 拉。此係藉由在信號BLS高(圖丨3(E))之情形下經由隔離電 晶體482將預充電電路連接至位元線並在S(}S低(圖丨3(f)) 之情形下切斷到達之源極之NAND鏈來完成。藉由將信號 HHL接通至高(圖13(B))來耦合入預充電/鉗位電路以%見圖 10)。以此方式,將開始上拉位元線(例如,圖13(H1)及 13(Π))。在位元線已充電接近於其目標值時,位元線預充 電之相位(2)將開始。在相位(2)中,預充電繼續,但單元 耦合至位元線以允許位元線電壓將在感測條件下穩定。第 一子循環通過期間之總預充電週期係由一預充電週期 表示。 此其中位元線初始自單元去耦合之實施例僅在允許耦合 之後其不招致一使位元線電壓穩定之長的等待週期時較 仏。換言之,若在其中位元線正好在位元線預充電操作開 始時耦合至單元之情形下等待週期短於穩定時間,則此實 鉍例係較佳。否則,具有其中不實施相位(〗)且位元線預充 電僅以其中位元線比對單元之傳導電流而預充電之相位(2) 開始之另一實施例將更佳。 感測在相位(3)中發生。如先前所述,在第一感測子循 133985.doc •30- 200921679Each sensing module 480 is coupled to a corresponding unit via a bit line 36. For example, a sensing module 480 senses the conduction current source-drain current of a memory cell. The conduction current flows from the sensing module through the bit ^36 into the memory unit 1G, and flows out of the source 14 through the source line ^J to the ground. In an integrated circuit chip, the secrets of the cells in a cell array are all connected as a branch of the source line 34, and the source line 34 is connected to an external ground of the memory chip. Solder pads (eg, Vss pads). Even when a metal strip is used to reduce the resistance of the source line, a finite resistance R is maintained between the source electrode of a memory cell and the ground pad. Usually, the ground loop resistance is about 5 〇 〇hm. For the entire memory page being sensed in parallel, the total current flowing through the source line is the sum of all conducted currents, that is, ^ corp..., lp-like, σ, mother-s-resonant unit Each has a conduction current that is dependent on the amount of charge programmed into its charge storage element. For a given control gate voltage of one of the memory cells, a small amount of charge will produce a relatively high conduction current (see Figure 3). When there is a finite resistance between the source electrode of a memory cell and the ground pad, the voltage drop across the resistor is given by Vcirop^Z'rorR. For example, if the 24,000 bit lines each having a current of 0.25 μA are simultaneously discharged, the source line voltage drop will be equal to 24 turns χ〇 25 μΑ / each χ 50 ohms to 0.3 volts. When sensing the threshold voltage of the memory cell, this source line bias will result in a sensing error of 0.45 volts, assuming that the body effect causes the source voltage to rise by 0, 3 V causing the threshold voltage to rise 〇 45 v. Figure 12B illustrates the error in the threshold voltage level of the memory cell 131985.doc -26- 200921679 caused by a source line voltage drop. The threshold voltage VT supplied to the control gate 30 of the memory cell 1 is relative to GND. However, the effective voltage vT of the memory cell is controlled by the voltage difference between the gate 30 and the source 14. There is a difference between the supplied ντ and the effective % - about i 5xvdrcp (ignoring the effect of a smaller voltage drop from the source 14 to the source line). When 纟 senses the threshold voltage of these memories, the ν" or source line bias will cause a sensing error of (for example, 0.45 volts). This bias cannot be easily removed because of The data (4) 'is dependent on the memory state of the memory cells of the page. The source load and the power saving technique using the bit line lock have been applied by U et al. on March 16, 2005. A power saving technique is disclosed in U.S. Patent Application Serial No. 1 1/083, No. 5, the entire disclosure of which is hereby incorporated by reference. The manner is incorporated herein. In particular, a read or program verify operation includes one or more sensing cycles corresponding to one or more demarcation threshold voltages to determine which of a plurality of possible memory states per memory cell Among them. Each sensing cycle is used to sense with respect to a threshold threshold voltage and process a memory cell page in parallel. A sensing loop typically contains more than P passes or sub-cycles to resolve the memory state of all cells in the page. In one aspect, a first pass or sub-cycle senses and identifies as many memory cells as possible with the highest conduction current in the page. This will minimize the presence of the high current cells during a subsequent sub-cycle. And caused the error of 131985.doc -27- 200921679 7 sensing. Since the cells have been read, their conduction current is turned off to save power. The cells are disconnected by grounding their associated bit lines such that there is substantially no potential difference across the source and drain of each cell. In a subsequent pass or sub-cycle, the remaining memory cells of the page will be sensed again in parallel with reduced interference from the higher current cells. Thus, at least two sense passes 35 are implemented by distinguishing between two adjacent memory states with respect to each threshold threshold voltage. The first pass or sub-cycle identifies a high current unit having a threshold voltage below the demarcation level. The second pass or sub-cycle is repeated after the bit line of the high current cell is locked to a ground potential to turn it off. Figure 10 also shows that one of the bit lines 36 pull-down circuit 55 〇 ' pull-down circuit 55 启动 is activated when both the signal INV and the other signal GRS are high. The pull-down circuit 55 is preferably formed by a first NMOS transistor 55 串联 in series with a second n transistor 552. Signals INV and GRS will turn on transistors 55A and 552, respectively, when both are high. This will pull the sense node 48 1 and also the connected bit line 36 down to the ground voltage. Regardless of the control gate voltage, this will suppress the flow of conduction current in the memory cell 1' because there will be no voltage difference between its source and drain. The effective 'signal GRS can be considered by the state machine as a control signal to enable (GRS = high) or disable (GRS = low) bit line lock. Fig. 13(Α)·13(1) is a timing diagram of the sensing through the bit line lock 2 . It should be understood that there will generally be a sensing cycle % for determining whether each cell in the page has a threshold voltage below - above or above the threshold level, the threshold threshold is used to distinguish Two memory states. For each-sensing cycle, there will be 2 sensing passes relative to the mother of the reference threshold voltage supported by the memory 131985.doc •28· 200921679. The specific map i3(A)-13(J) is a timing diagram for controlling the signals of the sensing module shown in Fig. 1A. The overall scheme senses the memory cell pages in parallel with respect to the established reference threshold or reference conduction current. As described above, by determining a conduction current of a memory cell with respect to a reference current, the threshold (4) in the cell is sensed relative to a threshold threshold level - having - lower than the boundary The limit current limit unit will cause its conduction current to be higher than the reference current. Therefore, if the sensing cycle performs the next threshold threshold in a step-up sequence, each sensing cycle will distinguish between cells having conduction currents lower than the reference conduction current. U.S. Patent No. 7,196,931 discloses a method of reducing the source bias error by sensing a loop by _2. The 2 pass sensing cycle has a first pass identifying the cells having conduction currents that are substantially higher than the reference current. The relative reference current measurements are performed in the second pass after the cells are identified and disconnected in the absence of high current cell interference. Each of the sensing relative to a reference threshold voltage thus includes at least two sub-cycles of phase (1)-(4) and phase (5)-(9), respectively, wherein each sub-cycle is When the memory cell page is sensed in parallel, it passes through. Each of the sense sub-cycles requires a setting to set the word line and bit line to the appropriate voltage before sensing can occur. This is done by a precharge operation. The pre-charge operation of the first sub-cycle is between phase (丨)_(2) and the pre-charge operation of the second cycle is between phases. Figure 13 (A) shows the timing of precharging of the selected word line. If the sensing system is at the threshold voltage level of vT1, the word line begins to be precharged to this voltage level. Looking at the word line relative to the bit line <RC delay rc delay, the pre-charge of the sub-line can begin earlier than the pre-charging of the bit line. Precharging of the bit lines can occur with or without the memory cells coupled to the bit lines. As previously described, in one embodiment, the cells are initially decoupled from the bit line such that their drain current does not prevent the bit line from being pulled up. This is done by connecting the precharge circuit to the bit line via the isolation transistor 482 with the signal BLS high (Fig. 3(E)) and at S(}S low (Fig. 3(f)). This is done by cutting off the NAND chain that reaches the source. By turning the signal HHL high (Figure 13(B)), it is coupled into the precharge/clamp circuit as shown in Figure 10). In this way, the pull-up bit line (eg, Figures 13 (H1) and 13 (Π)) will begin to be pulled up. The bit line precharge phase (2) will begin when the bit line has been charged close to its target value. In phase (2), precharge continues, but the cell is coupled to the bit line to allow the bit line voltage to be stable under sensing conditions. The total precharge period during the first sub-cycle pass is represented by a precharge period. This embodiment in which the bit line is initially self-decoupled by the cell is only awkward after allowing the coupling to not cause a long wait period for the bit line voltage to stabilize. In other words, this embodiment is preferred if the wait period is shorter than the settling time in the case where the bit line is coupled to the cell at the beginning of the bit line precharge operation. Otherwise, it would be preferable to have another embodiment in which the phase (") is not implemented and the bit line precharge is only started with the phase (2) in which the bit line is precharged with respect to the conduction current of the cell. Sensing occurs in phase (3). As previously described, in the first sensor, 133985.doc •30- 200921679

環中’所識別高電流單元。因此,感測係相對於一參考臨 限,該參考臨限可處於_將自彼用於下—感測子循環中之 邊際。換言之,第-子猶環可使用—處於—高於下一子循 環之電流之裕量之分界電流。在—個實施例中,此係藉由 縮短感測模組彻(見圖1G)之單元電流鑑別器65Q(見圖iib) 中之電容II 652之放電時間來完成。信號祖控制將預充 電電路搞合或去耗合至SEN節點之電晶體632(見圖1〇)且因 而單兀電流鑑別器650。另一方面,信號XXL控制將位元 線自SEN節點耦合或去耦合之電晶體63〇。在相位(3)開始 時,信號HHL變為低(圖13(B)),從而終止預充電且該單元 之傳導電流將使電容器652放電。放電週期之結束係藉由 在相位(3)結束時為低的xxl加以控制,從而藉由將位元線 自SEN節點去耦合來切斷電流。由圖丨丨b中所示之單元電 流鑑別器6 5 0可見,一將根據其分界之分界電流位準與放 電時間有關,一較長放電時間產生一較小分界電流位準。 在相位(4)中’接著相對於p電晶體656(見圖11B)之臨限 電壓比較經放電之電容器之電壓,且由選通信號STB鎖存 該結果。接著藉由縮短相位(3)中之感測週期來完成上述增 加裕量。以此方式,將僅使最高電流能夠在縮短之週期内 使電容器放電以斷開P電晶體656。 在第一子循環已識別高電流單元之後,接著在下一感測 之前將其鎖存並斷開。此係藉助具有一為INV=高之所感 測之結果之彼等高電流單元來完成。舉例而言,在圖1 3 中,耦合至位元線BL1之單元具有一約為120 nA之低於分 131985.doc •31 - 200921679 界臨限電流之傳導電流(見圖13(H1))。此相對較小電流將 能夠充分地使電容器652放電以減小SEN處之電麼(見圖 ΠΒ)以接通p電晶體656,以便在斷定選通stb信號時,信 號INV(圖13(H2)甲顯示為游!)不上拉至冑。因此,具有 相對較小電流之單元將具有INV=:低。High current unit identified in the ring. Thus, the sensing system is relative to a reference threshold that can be used at the margin of the sub-sensing sub-cycle. In other words, the first-sub-judging can use - a dividing current that is higher than the margin of the current of the next sub-cycle. In one embodiment, this is accomplished by shortening the discharge time of capacitor II 652 in the cell current discriminator 65Q (see Figure iib) of the sensing module (see Figure 1G). The signal master control combines or decouples the pre-charging circuit to the transistor 632 of the SEN node (see Figure 1A) and thus the current discriminator 650. On the other hand, the signal XXL controls the transistor 63 耦合 that couples or decouples the bit line from the SEN node. At the beginning of phase (3), signal HHL goes low (Fig. 13(B)), thereby terminating precharge and the conduction current of the cell will discharge capacitor 652. The end of the discharge period is controlled by xxl which is low at the end of phase (3), thereby cutting off the current by decoupling the bit line from the SEN node. As can be seen from the cell current discriminator 65 5 shown in Figure b, a demarcation current level according to its demarcation is related to the discharge time, and a longer discharge time produces a smaller demarcation current level. The voltage in the discharged capacitor is then compared in phase (4) with respect to the threshold voltage of p transistor 656 (see Fig. 11B), and the result is latched by strobe signal STB. This increase margin is then accomplished by shortening the sensing period in phase (3). In this manner, only the highest current will be able to discharge the capacitor to break the P transistor 656 during the shortened period. After the high current unit has been identified in the first sub-cycle, it is then latched and turned off before the next sensing. This is done by means of a high current unit having a result of sensing the INV = high. For example, in Figure 13, the cell coupled to bit line BL1 has a conduction current of approximately 120 nA below the threshold of 131985.doc • 31 - 200921679 (see Figure 13 (H1)). . This relatively small current will be sufficient to discharge capacitor 652 to reduce the charge at SEN (see Figure ΠΒ) to turn on p-transistor 656 so that when asserting the strobed stb signal, signal INV (Figure 13 (H2) ) A shows up as a swim!) Do not pull up to 胄. Therefore, a cell with a relatively small current will have INV = low.

方面,具有相對較大電流(例如,大於3〇〇 nA 另 見 圖13(11))之單元將具有鎖存於高之信號卿(圖叫⑺中顯 示為INV2)。&用於啓動圖1〇中所示之下拉電路55〇。在藉 由GRS信號高(圖13(J))啓用下拉電路時,接著只要爾為 高’下拉電路550即將經由所啓用之隔離電晶體482(見圖 剛)將位元線拉向接地。因此,感_具有相對較大電 流之單元之位元線將敎至接地,從而關閉彼等單元。 在由相位(5) (9)代表之第:感測通過或下__感測子循環 中,該過程類似於第一子循環。預充電週期702在相位(5)_ ⑷中發生。相位⑺中之感測在位元線中之電壓已藉助衰 減至某—不重要的值之位移電流穩定之後發生。選通及鎖 ::::⑻處發生且在第一個通過中遺漏之任何其他高電 ==似於先前相位⑷中圖13⑽中所示㈣ =反:接地。在相位⑼中,以實質上係信號 流排傳輸S E N形式之所感測之結果將經由讀出匯 等Π::元線鎖定至接地來關閉與電流感測無關之彼 寻早凡將有助於減小涵雷、、ώ · . a ^ 處。首先,此將、 2A)。此具有兩個益 、即功率。其次,在Vdrop隨著"or減小時 131985.doc -32- 200921679 此將減小源極之(CLSRC)接地迴路偏壓誤差。因此,現有 感測技術已相對於每個記憶體狀態實施此兩通過感測。 圖14不意性地圖解說明在圖8中所示之8狀態記憶體上應 用現有兩通過感測方案之_實例。該8狀態記憶體係由至 少7個分界臨限電壓位準(亦即,Da、Db、dc、Dd、De、In the aspect, a cell having a relatively large current (e.g., greater than 3 〇〇 nA or see Fig. 13 (11)) will have a signal latched at a high level (shown as INV2 in (7)). & is used to start the pull-down circuit 55〇 shown in FIG. When the pull-down circuit is enabled by the GRS signal high (Fig. 13(J)), then the pull-down circuit 550 is about to pull the bit line to ground via the enabled isolation transistor 482 (see figure just). Therefore, the sense lines of the cells with relatively large currents will be grounded to turn off their cells. In the first: sense pass or down __ sense sub-cycle represented by phase (5) (9), the process is similar to the first sub-cycle. The precharge cycle 702 occurs in phase (5)_(4). The sensed voltage in phase (7) has occurred after the displacement current has been stabilized by attenuating to a certain - unimportant value. Gating and Locking Any other high power that occurs at ::::(8) and is missing in the first pass == as shown in Figure 13 (10) in the previous phase (4) (4) = Reverse: Ground. In phase (9), the result of sensing the SEN form in a substantially signal stream will be assisted by reading the sink or the like: the meta-line is locked to ground to turn off the current sense independent of the current sense. Reduce the culvert, ώ · . a ^. First of all, this will, 2A). This has two benefits, power. Second, when Vdrop decreases with "or 131985.doc -32- 200921679 this will reduce the source (CLSRC) ground loop bias error. Therefore, existing sensing techniques have implemented this two-pass sensing with respect to each memory state. Figure 14 is an unintentional illustration of an example of applying an existing two pass sensing scheme to the eight state memory shown in Figure 8. The 8-state memory system has at least 7 boundary threshold voltage levels (i.e., Da, Db, dc, Dd, De,

Dd Dg)分界。因此’將存在至少一樣多的感測循環,各 自與該等分界臨限電壓位準中之一者相關聯。舉例而言, DA在記憶體狀態”Gr”與”a”之間分界,%在記憶體狀態"A" 與B之間分界等等。在每一感測循環期間,分界臨限電 壓位準將施加至選定字線。 每一感測循環將進一步具有兩通過。第一個通過包含一 後跟所偵測之鬲電流位元線之一鎖定之"預感測”。該預感 測將感測並識別具有低於所施加之分界臨限電壓位準之臨 限電壓之兩電流狀態。與該等所識別之高電流單元相關聯 之位元線將藉由鎖存至接地而鎖定。在移除高電流狀態之 情形下,第二通過將能夠更準確地感測。同樣,將適當地 識別且亦鎖定在第一個通過中未所識別之任何高電流狀態 以此方式,將準確地感測該等單元以使資料低於或高於所 細加之分界臨限電廢位準。 為區分所有可能的記憶體狀態,依次相對於每個分界臨 限位準感測記憶體單元頁。在分界臨限位準向該等單元之 臨限窗中之一較高值移動時,該頁中之更多單元將很可能 被識別為具有較高電流(具有低於分界臨限電壓位準之臨 限電廢)且因而將鎖定該頁之更多位元線。 131985.doc •33- 200921679 如先前所述,位元線鎖定係藉由圖丨0中斯_ 汀不之下拉電路 550來完成。在現有兩通過感測方案中, ^ Τ下拉電路5〇〇始終 準備只要經鎖存之信號INV為高即將被下拉。在圖ι〇中所 示之下拉電路550中,藉由使信號GRS始終高以使η電晶體 552始終提供一到達接地之連接來啓用該電路。 由於位元線鎖定而引起之效能及功率問題 美國專利第7,196,93 1號已指示藉由使用位元線鎖定來關 、 %已經感測或在電流感測中不再有關之單元,該兩通過感 測方案幫助限制所涉及之最大電流且亦在第二通過中由於 減小之源極之接地迴路偏壓誤差而提供更準確之感測。然 而,任何優點皆被由於感測通過之多樣性及由位元線鎖定 操作所產生之雜訊所引起之效能減小抵消。 記憶體陣列中之位元線中之接地選擇性位元線由於該等 位元線之間之電容而具有相互作用。位元線之間之電容隨 著積體電路密度越來越高而變得日益重要。對於具有先前 ) 提及之所謂的"所有位元線"("ABL")架構之記憶體而言, 位元線至位元線電容可甚至更高。一所有位元線架構中之 頁係由沿一列之相鄰串記憶體單元形成。若記憶體平面沿 位元線方向較長,則ABL之位元線至位元線電容可更高。 奴而。,在ABL及習用架構兩者中,位元線至鄰近位元 線距離相同。在習用情形中,一半位元線預充電而其最近 的鄰居保持接地,此係所有位元線至位元線串擾電容正面 臨之最差的情形情節。在ABL中,所有位元線皆一同充 電’但接著在不同時間放電。 131985.doc • 34- 200921679 由於位7G線(及字線)起電容性負載的作用,故在一位元 線正被預充電或放電時,存在兩個與位元線鎖定方案相關 聯之不期望之效應。 首先’許多位it線將鎖定在接地電位,而#他位元線正 在預充電期間拉向一較高電位。由於位元線至位元線電 容’其-般而言將較困難且與將頁中之所有位元線拉在一 起而不面臨某些位元線將接地之情形相比較將必須消耗更 多功率來在一錨定之接地之位元線背景中預充電位元線。 其次,一交流("AC")位移電流將初始流動並最终隨著位 元線變得充電至所施加之電屢而衰減至零。衰減時間係位 :線之RC常數之一函數,其中c為有效電容。由於感測一 單元係為確定其直流(,,DC”)傳導電流所必不可少,故位元 線中之準確之感測可僅在AC位移電流已衰退之後開始。凡 正感測之位元線中流動之AC位移電流之一實質部分可 視為來自充電位元線至最近之位元線電容,乃因大多數每 -位元線之總電容係由每一位元線至駐於其側上之其兩個 鄰居之電容组成。每-位元線或每一電極之電容皆等於電 極至所有其鄰居之電容之總和。若減去至所有鄰居之所有 電容’則沒有什麼留下。每一位元線之總電容之約9〇%係 至其第一、第二及第三最近的鄰居。此為上方或下方之屏 留下略超過10%之位元線電容。 若所有位70線皆同時充電且稍後同時放電,則每—位元 線之有效電容僅約每-位元線之總電容之1G%。若其所: 皆同時充電但在各種感測操作期間其在各個時間下降至接 131985.doc •35· 200921679Dd Dg) Demarcation. Thus there will be at least as many sensing cycles, each associated with one of the boundary threshold voltage levels. For example, DA is demarcated between the memory states "Gr" and "a", and % is demarcated between the memory state "A" and B, and so on. During each sensing cycle, a threshold threshold voltage level is applied to the selected word line. Each sensing cycle will have two more passes. The first is a "pre-sensing" that is locked by one of the 鬲 current bit lines detected by the follower. The pre-sensing will sense and identify a threshold that is lower than the applied threshold voltage level. Two current states of voltage. The bit line associated with the identified high current unit will be locked by latching to ground. In the case of removing the high current state, the second pass will be more accurate Similarly, any high current state that is not identified in the first pass will be properly identified and locked in such a way that the cells will be accurately sensed so that the data is below or above the finest boundary In order to distinguish all possible memory states, the memory cell pages are sequentially sensed relative to each threshold threshold. One of the threshold windows of the boundaries is bounded by the boundary threshold. When a high value is moved, more of the cells in the page will most likely be identified as having a higher current (with a threshold charge below the threshold threshold voltage level) and thus will lock more bit lines of the page. 131985.doc •33- 200921679 as previously The bit line lock is completed by the _ 不 不 pulldown circuit 550. In the existing two pass sensing scheme, the ^ Τ pull-down circuit 5 〇〇 is always prepared as long as the latched signal INV is high It is about to be pulled down. In the pull-down circuit 550 shown in Figure ι, the circuit is enabled by having the signal GRS always high so that the η transistor 552 always provides a connection to ground. This is caused by the bit line lock. Performance and Power Issues U.S. Patent No. 7,196,93 1 has indicated that by means of bit line locking, % has been sensed or no longer relevant in current sensing, the two pass sensing schemes help limit the involved The maximum current also provides a more accurate sensing in the second pass due to the reduced ground loop bias error of the source. However, any advantage is due to the diversity of sensing passes and the bit line locking operation The performance reduction caused by the generated noise is offset. The grounded selective bit lines in the bit lines in the memory array have interactions due to the capacitance between the bit lines. Capacitance The density of integrated circuits is becoming more and more important and becoming more and more important. For the memory of the so-called "all-line"("ABL") architecture mentioned in the previous), the bit line to the bit The line capacitance can be even higher. A page in all bit line architectures is formed by adjacent column memory cells along a column. If the memory plane is longer along the bit line direction, the ABL bit line to the bit The line capacitance can be higher. In the ABL and the conventional architecture, the distance from the bit line to the adjacent bit line is the same. In the conventional case, half of the bit line is precharged and its nearest neighbor remains grounded. The worst case scenario in which all bit-to-bit line crosstalk capacitors are facing. In ABL, all bit lines are charged together' but then discharged at different times. 131985.doc • 34- 200921679 Since the 7G line (and word line) acts as a capacitive load, there are two associated with the bit line locking scheme when one bit line is being precharged or discharged. The effect of expectations. First, many of the bit lines will be locked to ground potential, while the # bit line is pulled to a higher potential during pre-charging. Since bit-to-bit line capacitance 'is generally more difficult and would have to consume more than pulling all the bit lines in the page without facing some bit lines to be grounded The power is to precharge the bit line in the background of the bit line of the anchored ground. Second, an alternating current ("AC") displacement current will initially flow and eventually decay to zero as the bit line becomes charged to the applied power. Decay time line: A function of the RC constant of the line, where c is the effective capacitance. Since sensing a unit is essential to determine its DC (, DC) conduction current, accurate sensing in the bit line can only begin after the AC displacement current has decayed. The substantial part of the AC displacement current flowing in the line can be regarded as the capacitance from the charging bit line to the nearest bit line line, because the total capacitance of most per-bit lines is from each bit line to the resident The capacitance of its two neighbors on the side. The capacitance of each bit line or each electrode is equal to the sum of the capacitances of the electrodes to all its neighbors. If you subtract all the capacitances of all neighbors, there is nothing left. About 9〇% of the total capacitance of each bit line is tied to its first, second, and third nearest neighbors. This leaves a slightly more than 10% bit line capacitance for the upper or lower screen. When the 70 lines are simultaneously charged and discharged at the same time, the effective capacitance of each bit line is only about 1 G% of the total capacitance of each bit line. If it is: both are charged simultaneously but during various sensing operations Every time dropped to 131985.doc •35· 200921679

地,則將必箔杵AS 、^擊之有效位元線電容甚高並可端視是否相 對於^鄰居同時或在不同的時間鎖定位元線而每一位元線 有:灸化由於位元線鎖定將存在許多不同的時機,故在 貞定,·、鄰居時同時鎖定—位元線之機會可頗低。 在同時2充電所有位元線時,所消耗之能量為Cxy2,其中 CxV係儲存作為電容器電極之間之電介質中之電場, X Cxv係轉換成跨越遞送該能量之電源之内部電 阻燃燒之熱之能量。此第二項不相依於電壓/功率源之内 4電阻值。只要所有位元線皆一致地充電,所有該等表達 式中之C即係有效c,其僅係如先前所解釋之每一位元線 之、’心電令之約1 〇%。然而,位元線在與其鄰居不同之時間 鎖定時(此相當經常地出現),則電源無須再向被下拉接地 之位元線遞送任何能量。其僅須向將維持在位元線電壓但 由被迫接地之鄰居電容性地下拉的鄰近位元線供應能量。 對於將維持在其位元線電壓而其鄰居被下拉至接地之位元 線而5,整個位元線至位元線電容被佔用。正被接地之位 元線將具有儲存於其周圍電介質之電場中之轉換成沿到達 接地之電阻性路徑耗散的熱能量。 圖1 5圖解說明三個毗鄰位元線及其之間之電容性耦合效 應。一記憶體單7C 10-0具有兩個毗鄰記憶體單元⑺—丨及⑺一 2。類似地,三個毗鄰位元線36_〇、“一及刊」分別耦合至 該三個記憶體單元。每一位元線之自身電容由其到達所有 其他電極(例如,其一對第一最近的鄰居、其一對第二最 近的鄰居、其一對第三最近的鄰居等等)之電容之總和組 I31985.doc -36 - 200921679 成,除其到達駐於其上方或下方之電極之電容之外。圖i5 並未描綠上述所有電容,而是描繪具有最大重要性之最大 電容。 接著可看出,由於各種電容,可能存在各種電流支路。 特疋而S,由於每一位元線自身電容引起之電流將形成: IBLO - C8L0, d/dt (VBL0 - VBLI) + Cbl〇2 d/dt (VBL〇 - VBL2) 忽略第二、第三及其他鄰居之效應及所關注位元線上方 〇 或下方之層中之電極的效應。若所有位元線皆一同充電, 則以上表達式為零,且位移電流將係由對應於所關注位元 線與在其上方或下方之層中之電極之間的電容達到使該等 層之電壓不與所關注位元線之電壓協作地移動之程度的忽 略項所造成。 以上給出之單元電流為一近似值,乃因其僅包含來自毗 鄰位兀線之影響。一般而言,對於位元線BL0而言,亦將 存在由於左側的非毗鄰位元線而引起之電容Cbl〇3及由於右 (J 側的非毗鄰位元線而引起之電容CBL〇4。類似地,在非毗鄰 位7L線BL1與BL2之間將存在一共同電?Cbli2。該等電容 導致相依於跨越每一電容器之變化電壓之位移電流。 總位元線電流接著由位移電流與傳導電流之總和組成。 感測安培必須提供進入位元線並通過單元且接著進入接地 之正傳導電流。同樣,在其鄰居正被下拉至接地之位元線 上’感測安培必須提供額外正電流以抗擊所導致之位移電 流。被下拉至接地之位元線僅接地且不需要來自感測安培 I31985.doc -37- 200921679 之功率或任何其他供應來使一電極到達接地。 就充電交又耦合電容而言,位移電流將相依於位元線之 間之電壓差之改變速率。電壓差之改變速率可來自位元線 與其鄰居之間之不同充電及放電速率。 在預充電期間,如先前所述,耦合至更多導電單元之位 元線將需要更多淨電流用於電壓充電且可因而與一具有一 較少導電單元之鄰居相比較充電較i曼。因此,在位元線及 其鄰居具有類似的記憶體狀態且因而類似的導電電流時, 其將皆以類似的速率充電並在任一既定時刻具有類似的電 壓。在此情形下,跨越交叉耦合電容之電壓差將相對較小 且相關聯之位移電流將同樣如此。同# ’藉由降低位元線 電壓,不僅需要較少能量來充電所涉及之電容,且亦線性 地減小最大傳導電流。期望之效應係一具有一固定強度之 感測放大器可輕易維持所有位元線之充電升載率,儘管某 些位元線連接至不導電單元而其他位元線連接至導電單元 之事實。 在車乂佳實施例中’耗合至複數個記憶體單元之複數個位 元線之位元線電壓經控制以使每4鄰對位元線之間之電 麼差大致不相依於正感測其傳導電流之時間。在施加了此 條件時’由於各種位元線電容所引起之所有電流被略去, 乃因其皆相依於一隨時間變化之電壓差。因此,根據上方 方知式&於⑽%⑽邱⑽卜0,故自位元線感測到 之電流與單元之電流㈣,例如— 最大位移電流將在與正感測之位元線相關聯之單元及其 131985.doc -38- 200921679 鄰居具有不同的記憶體狀態時出;見。舉例而言,正感測之 : 立元線:其鄰居輕合至高度導電之單元時搞合至一不導電 隱體單兀。一般而言,在位移電流極其衰減生命時間範 圍中將存纟$佈。此意味著’預充電恢復操作彼處在一 預二週期期間發生’該預定週期延伸超過最差情形恢復次 "在位元線將被視為可穩定以準確之感測之前位移電 流必須已衰減至一預定位準。 在2通過感測方牵中, 茶甲右在第一個通過結束時鎖定某些 毗鄰位7L線,則第二通過之預充電週期進一步增加。 在位元線鎖定方案中,所識別之高電流單元正好在為第二 通過預充電之前使其位元線鎖存至接地。在某些位元線被 :速地自—先前預充電之位準拉至接地電位而其他位= 大致維持在預充電之電位之情形下,電麼差之改 極端。此將在將發生下—感測之情形下在位元線中導致— 電流。該預充電週期必Μ夠長以使所有位移電 鎖 減掉此’藉助該位元線 鎖疋方案’在母個鎖定操作之後,必須提供更 Μ線電壓在感測可發生之前穩定。再次參照圖14,每= 測之間之此延遲在每一相對於-分界臨限位 盾衣甲出現兩_人。對於多級記憶體而言 :-步在每-分界臨限位準處複合。舉例而言,對二8狀 ’由於位元線鎖定效將存在十 (2個购期用於將區分8個狀態之七個讀取位準)= 致使嚴重的效能降格。 攸而 131985.doc •39· 200921679 藉助f擇性啓用位元線鎖定進行感測 根據本發明之— 時,將—位 〜…,並行地感测一記憶體單元頁 ^ a 線接地之情形(其係一位元線接地以關μ 一 〜-預定之電流位準之記憶體單才: 方式,將介邻蛊、土 ,啤小至最小。以此 關閉高電产單:、㈣電流消耗之一既定預算而跳過識別及 感測=;Γ之額外感測子循環,且以-選擇性數目之 = 壓進行感測將導致偵測對應於接通單元之Ground, the effective bit line capacitance of the foil 杵AS and ^ is very high and can be viewed whether the bit line is locked at the same time or at different times with respect to the neighbors. Each bit line has: moxibustion due to bit There are many different opportunities for meta-line locking, so the chances of locking the bit line at the same time as the neighbors can be quite low. When all the bit lines are charged at the same time 2, the energy consumed is Cxy2, wherein CxV is stored as an electric field in the dielectric between the capacitor electrodes, and X Cxv is converted into heat of internal resistance combustion across the power source that delivers the energy. energy. This second term does not depend on the 4 resistance values within the voltage/power source. As long as all of the bit lines are uniformly charged, C in all of these expressions is valid c, which is only about 1% of the 'electrocardiogram' for each bit line as previously explained. However, when the bit line is locked at a different time than its neighbor (which occurs quite often), the power supply no longer has to deliver any energy to the bit line that is pulled down to ground. It only has to supply energy to adjacent bit lines that will sustain the bit line voltage but are capacitively pulled down by the neighbors that are forced to ground. For a bit line that will remain at its bit line voltage and its neighbor is pulled down to ground, the entire bit line to bit line capacitance is occupied. The bit line being grounded converts the electric field stored in the dielectric around it into thermal energy that is dissipated along the resistive path to ground. Figure 15 illustrates three adjacent bit lines and their capacitive coupling effects. A memory bank 7C 10-0 has two adjacent memory cells (7) - 丨 and (7) - 2. Similarly, three adjacent bit lines 36_〇, "一一刊" are respectively coupled to the three memory cells. The sum of the capacitances of each bit line's own capacitance from all other electrodes (eg, its pair of first nearest neighbors, its pair of second nearest neighbors, its pair of third nearest neighbors, etc.) Group I31985.doc -36 - 200921679, except that it reaches the capacitance of the electrode above or below it. Figure i5 does not depict all of the above capacitors, but rather depicts the largest capacitance of greatest importance. It can then be seen that various current branches may exist due to various capacitances. In particular, the current due to the capacitance of each bit line is formed: IBLO - C8L0, d/dt (VBL0 - VBLI) + Cbl〇2 d/dt (VBL〇- VBL2) Ignore the second and third And the effect of other neighbors and the effect of the electrodes in the layer above or below the bit line of interest. If all of the bit lines are charged together, the above expression is zero, and the displacement current will be such that the capacitance between the electrodes corresponding to the bit line of interest and the layer above or below it reaches The voltage is not caused by an ignoring of the degree of movement of the voltage of the bit line of interest. The cell current given above is an approximation because it only contains the effects from the adjacent bit 兀 line. In general, for the bit line BL0, there will also be a capacitance Cbl〇3 due to the non-adjacent bit line on the left side and a capacitance CBL〇4 due to the right side (the non-adjacent bit line on the J side). Similarly, there will be a common charge Cbli2 between non-adjacent bits 7L lines BL1 and BL2. These capacitances cause displacement currents that are dependent on varying voltages across each capacitor. The total bit line current is then displaced by current and conduction. The sum of the currents. The sense amp must provide a positive conduction current that enters the bit line and passes through the cell and then into ground. Similarly, the sense amp must provide additional positive current on the bit line whose neighbor is being pulled down to ground. The displacement current caused by the impact. The bit line pulled down to ground is only grounded and does not require power from sensing amps I31985.doc -37- 200921679 or any other supply to bring an electrode to ground. In other words, the displacement current will depend on the rate of change of the voltage difference between the bit lines. The rate of change of the voltage difference can come from different charging and discharging between the bit line and its neighbors. During pre-charging, as previously described, bit lines coupled to more conductive cells will require more net current for voltage charging and may thus be charged compared to a neighbor with one less conductive cell. Thus, when a bit line and its neighbors have similar memory states and thus similar conduction currents, they will all charge at a similar rate and have similar voltages at any given time. In this case, crossover The voltage difference of the coupling capacitor will be relatively small and the associated displacement current will be the same. By reducing the bit line voltage, not only does it require less energy to charge the capacitor involved, but also linearly reduces the maximum conduction. Current. The desired effect is that a sense amplifier with a fixed strength can easily maintain the charge boost rate of all bit lines, despite the fact that some bit lines are connected to non-conducting cells and other bit lines are connected to conductive cells. In the embodiment of the vehicle, the voltage of the bit line of the plurality of bit lines that are consumed to the plurality of memory cells is controlled so that each of the four adjacent bit lines is between The difference is substantially independent of the time at which the conduction current is being sensed. When this condition is applied, 'all currents due to various bit line capacitances are omitted because they are all dependent on a time-varying voltage difference. Therefore, according to the above formula & (10)% (10) Qiu (10) Bu 0, the current sensed from the bit line and the current of the unit (4), for example - the maximum displacement current will be in the line with the positive sense The associated unit and its 131985.doc -38- 200921679 neighbors have different memory states; see. For example, positive sensing: 立元线: when its neighbors are lightly coupled to a highly conductive unit To a non-conducting hidden body unit. In general, the displacement current will be stored in the extremely attenuated life time range. This means that the 'precharge recovery operation occurs during a pre-second period'. Exceeding the worst case recovery times " The bit line will be considered stable and accurate to be sensed before the displacement current must have decayed to a predetermined level. In the 2 sense pass, the tea right locks some adjacent 7L lines at the end of the first pass, and the second pass precharge period is further increased. In the bit line locking scheme, the identified high current cell latches its bit line to ground just before the second pass precharge. In some cases where the bit line is quickly pulled from the previous pre-charge level to the ground potential and the other bits = approximately maintained at the pre-charge potential, the power difference is extreme. This will result in a current in the bit line in the event that the next-sensing will occur. The precharge period must be long enough for all of the displacement locks to be subtracted. By this bit line lock scheme, after the parent lock operation, a more twist line voltage must be provided to stabilize before sensing can occur. Referring again to Figure 14, this delay between each test occurs in two _ persons with respect to the armor of the shield. For multi-level memory: - Steps are compounded at the per-demarcation threshold. For example, there will be ten (2 purchase periods for the seven read levels that will distinguish 8 states) due to the bit line locking effect = causing severe performance degradation. 13 13 131985.doc •39· 200921679 by means of opt-in enabling bit line lock for sensing according to the invention - when - bit ~ ..., in parallel sensing a memory cell page ^ a line grounded ( It is a memory line that is grounded by a bit line to close the current level of μ~~ predetermined current level: way, the neighboring 蛊, soil, and beer are minimized. This turns off the high power production list: (4) current consumption One of the established budgets skips the identification and sensing =; additional sensing sub-cycles, and sensing with a selectable number of = will cause the detection to correspond to the connected unit

一此ΓΓ而以其他控制間極電麼進行感測將不導致任 匕位凡線關閉操作。通過庳 因減小感測子循…感測操作之效能 : 衣之數目且在關閉位元線時減小由於位元 ' 70線耦合而產生之雜訊而得以改良。在此上下文 中’位元線至位元線耦合係指鄰近全局位元Simultaneous sensing with other control poles will not result in any line closing operation. By reducing the efficiency of the sensing subsense... sensing operation: the number of clothing and the reduction of the noise generated by the bit line '70 line coupling when the bit line is turned off is improved. In this context, 'bit-to-bit line coupling refers to neighboring global bits.

電容性耦合。 + βI 在-較佳實施例中,位^線敎係藉由—能夠將該位元 線拉至接地之下拉電路來實施。該下拉電路包括兩個在位 兀線與接地之間串聯之通過閘極。該兩通過問極來自一 編閘極,其中一個通過問極由一下拉啓用或停用控制信 號加以控制且另一個通過閘極由是否感測到討論中之單元 具有一高於或低於一參考電流之電流加以控制。在停用下 拉電路時,不管所感測之結果如何位元線皆將不接地。在 啓用下拉電路時,在所感測之結果係來自—高電流記憶體 單元時位元線將被拉至接地。 圖16(A)-16(J)係控制併入選擇性位元線鎖定之感測模組 之操作之信號之定時圖。實質上,圖10中所示之感測模組 131985.doc •40、 200921679 4 8 0具有藉由控制信號G R S啓用或停用之位元線鎖定特 信號咖係由狀態機叫見圖⑽應”在信號⑽為 两時’啓用下拉電路55G(見圖1G)。相反地,在grs為低 時’ ^用下拉電路55〇。在彼態樣中,圖13(α)·ι3⑺中所 不之定時圖較早指代在在信號GRS始終為高(圖 形下啓用位元線鎖定之情形。另一方面,圖 /Λ =定時圖指代可在信號⑽為低(圖16(J))之情形下可 、擇!·生地停用位元線鎖定之情形。 =些實施例中,在啓用位㈣鎖定時,其先於一感測 T循衣以確定鎖定之高電流單元。另一方面,在選擇性地 2用位S線鎖;t時,亦將跳過其用於確定高電流單元之先 别感測子循環。 、圖16(Α)_16σ)係圖解說明選擇性地停用位元線鎖定期間 /刀別相對於兩個連續分界臨限位準之兩個感測循環之定時 :。因此,不像在啓用位元線鎖定之情形,每一 括一 1通過感測。 =相對於Klit過感測,將選定字線預充電至、 =位元線的預充電之前或與其同時發生。特定而言’ 2線預充電㈣發生在相位(1.5)_(16)中。相 :感測發生在位元線中之電壓已藉助衰減至某-不重要的 =之::位移電流穩定之後1通及鎖存發生在相位 鎖存為有低於VT1之臨限電覆之”高”電流單元具有 有鎖存虎1Nv’且具有高於VT1之臨限電麼之單元具 存為低之mv。在相位(1.9)中,以實f上係信號⑽ J31985.doc 41 200921679 之反轉之信號SEN形式之所感測之結果將經由讀出匯流排 傳輸出去。 自圖16(J)將可見,信號GRS對兩個感測循環而言皆低, 從而停用每一位元線之下拉電路55〇(圖1〇)不管所感測之 INV值如何。此意味著即使對於具有低於Vti(見圖ΐ6(ιι) 及16(12))之臨限電壓之高電流單元而言,其位元線將藉由 拉至接地而不被鎖定。Capacitive coupling. + βI In the preferred embodiment, the bit line is implemented by pulling the bit line to the ground pull-down circuit. The pull-down circuit includes two pass gates connected in series between the bit line and ground. The two pass gates are from a gate, one of which is controlled by a pull-up enable or disable control signal and the other pass-through is sensed by whether the cell in question has a higher or lower than one The current of the reference current is controlled. When the pull-down circuit is disabled, the bit line will not be grounded regardless of the result of the sensing. When the pull-down circuit is enabled, the bit line will be pulled to ground when the sensed result is from the high current memory unit. 16(A)-16(J) are timing diagrams for controlling signals incorporating the operation of a selective bit line locked sensing module. In essence, the sensing module 131985.doc • 40, 200921679 4 8 0 shown in FIG. 10 has a bit line locking special signal enabled or disabled by the control signal GRS. The state machine is called as shown in the figure (10). The pull-down circuit 55G is enabled when the signal (10) is two (see Fig. 1G). Conversely, when the grs is low, the pull-down circuit 55 is used. In the aspect, the figure is not in the figure (α)·ι3(7). The timing diagram refers earlier to the case where the signal GRS is always high (the bit line lock is enabled under the graph. On the other hand, the graph / Λ = timing diagram refers to the signal (10) is low (Figure 16 (J)) In the case of the choice, the bit line lock is disabled. In some embodiments, when the bit (4) lock is enabled, it precedes a sense T to determine the locked high current unit. On the other hand, when selectively using the bit line lock; t, the first sensing sub-cycle for determining the high current unit will also be skipped. Figure 16 (Α)_16σ) illustrates the selective stop The timing of the two sensing cycles with the bit line locking period/knife relative to the two consecutive demarcation threshold levels: therefore, unlike the enabling bit line In the case of locking, each 1 is sensed. = Relative to Klit over sensing, the selected word line is precharged to, before or at the same time as the pre-charging of the bit line. In particular, '2-wire pre-charging (4) Occurs in phase (1.5)_(16). Phase: Sensing occurs in the bit line. The voltage has been attenuated to a certain - unimportant =:: Displacement current is stable after 1 pass and latch occurs in phase lock The "high" current unit with a threshold voltage lower than VT1 has a latched tiger 1Nv' and has a lower limit of VT1. The unit has a low mv. In phase (1.9), The result of the sensing of the signal SEN in the inverse of the signal (10) J31985.doc 41 200921679 will be transmitted via the readout bus. As can be seen from Figure 16(J), the signal GRS is applied to two sensing cycles. In terms of low, the deactivation circuit 55〇 (Fig. 1〇) of each bit line is deactivated regardless of the sensed INV value. This means that even if there is less than Vti (see Figure ΐ6(ιι) and 16( For the high current unit of 12)), the bit line will be locked by grounding without being locked.

為限制頁電流及源極偏壓誤差,使用位元線鎖定方案以 藉由將其位元線鎖定至接地以使其不再相關於將實施之下 -感測通過來關閉高電流單元。因此,現有感測技術已藉 助位元線鎖定相對於每個記憶體狀態該實施兩通過感測。 亦已解釋’兩通過感測之缺點係其可極大地影響感測效 能。實施兩通過感測將使感測通過之數目加倍,從而相對 於單通過感測將感測操作延長約雙重。實際上,由於需要 管理由將位元線鎖存至接地所產生之瞬時雜訊,該延=甚 長。如先前所述,引起一顯著位移電流量作為進入一正感 測其電流之位元線中之雜訊。該位移電流係由於鄰近位元 線之間的電壓因該等位元線之相互電容所致之速率改變而 引起。因Λ,充分延長感測循環之位元線預充電週期以等 待直至位移電流衰減掉。在此延遲之週期期間,該記憶體 二元搞合至位it線且因而在預充電電路上拉位元線㈣該 單元之傳導電流妨礙其之情形下消耗額外功率。 藉由不實施所偵測之高電流單元之位元線鎖定,僅 m過感測且此外位元線不拉至接地。雜訊將最小化且不 131985.doc •42- 200921679 需要延長位元線預充電週期。此藉由圖16中所示之Vt2感 測循環中之相位(2.5)之縮短之預充電週期702加以示意性 地圖解說明。事實上,已估計該等單元之全部接通週期皆 縮短,從而導致功率節省。 在根據本發明減小位元線鎖定之數目以改良感測效能之 情形下,更多高電流單元將維持在接通狀態以導致如先前 結合圖12Α及12Β所述之源極偏壓誤差。不管"⑽R之高位 準,一種最小化誤差之解決方法係將一盡可能接近記憶體 單之源極之節點作為參考來參考到達個別記憶體單元之 控制閘極及汲極之所有電壓。舉例而言,如圖12A及i2B 中所示,參考點可選取於源極線32處而非接地以最小化接 地迴路電阻。已在美國專利第7,17〇,784及7,173,854號及 Sekar等人於2007年4月24日申請之題為”c〇mpensating SOURCE VOLTAGE DROP in NON-VOLATILE STORAGE”之 美國專射請案第U/739,5()1中揭示用於最小化源極偏壓 誤差之技術,该等專利之全部揭示内容以引用方式併入本 文中。 預疋之電位準在一個感測一多狀態記憶體之實施方案 (”牽扯相對於夕個狀@中之每—者對記憶體單元頁實施 少個感測通過)中’僅以預定感測通過實施鎖定與識別為 高於預定之電流位準之記憶體單元相關聯之位元線之步 驟以此方式,位元線鎖定以減小總電流及源極偏壓誤差 之益處與由於更多子循環而引起之較長感測時間及所產、生 之雜訊衰退之較長等待時間之負效應保持平衡。藉由選擇 13I985.doc -43« 200921679 性地減小鎖定之數目,減輕導致較低感測效能及較高功率 消耗之該等負效應。 圖17A圖解說明用於在一多狀態感測操作之多通過中選 擇性地啓用位元線鎖定之一個例示性排程。該實例與圖8 中所示之8狀態記憶體有關。為解析所有可能的八個狀 態,將必須在至少7個感測循環中感測記憶體單元頁,每 一個循環具有一不同的分界臨限位準,例如Da*Db,…,等 等。在每一感測循環中實施兩通過感測。然而,與現有兩 通過感測(見圖14)之差別係將在第二通過期間停用位元線 鎖定操作。換言之,將每隔一感測通過跳過位元線鎖定操 作。特定而言,將在第二通過中之感測之後跳過位元線鎖 定刼作。在通過之數目與現有兩通過感測相同時,位元線 鎖定操作減小百分之50。在此情形下,在第二通過之後將 產生較少雜訊並允許下一循環中之第一個通過具有一縮短 之位元線預充電週期。若字線穩定時間具有與位元線至位 70線串擾恢復次數相同之量級,則此實施例不顯著增加效 能,乃因位元線鎖定及自其恢復可與在下一感測位準處穩 定之字線電壓同時發生。 圖1 7B圖解說明用於在—多狀態感測操作之多通過中選 擇性地啓用位元線鎖定之另—例示性排程。該實例亦與圖 8中所不之8狀態記憶體有關且類似於圖1 7 A中所示之實 例’除將在每隔—感測猶環中實施兩通過感測之外。在具 固通過之循環中,將僅實施感測操作且將跳過預感測 及位元線鎖定操作。在此情形下,通過之數目減小百分之 131985.doc -44 - 200921679 5 0且位元線鎖定操作之數目減小 曰分之75。 圖1 7C圖解說明一儲存僞隨機化之次、 頁。為確保一頁中高電流狀態之岣勻八佈料之5己憶體單兀 碼以跨越該頁以相對一致分佈將資料二=’、该頁較佳經編 憶體狀態中。在-較佳實施例中4 :於所有可能的記 看起來僞隨機地分佈於可能的記怜 遇订編碼以 隨狀態中。以 可基於所跳過之位元線鎖定操作之 万式, 目以統計方式仕叫_ # 憶體系統中之電流量。 飞估。t 〇己To limit page current and source bias errors, a bit line locking scheme is used to turn off the high current cells by locking their bit lines to ground so that they are no longer relevant for the implementation-sensing pass. Therefore, existing sensing techniques have implemented two pass sensing with respect to each memory state by means of bit line lock. It has also been explained that the shortcomings of the two-pass sensing are that they can greatly affect the sensing performance. Implementing two pass sensing will double the number of sense passes, thereby extending the sensing operation by a factor of two relative to single pass sensing. In fact, this delay = very long due to the need to manage the instantaneous noise generated by latching the bit line to ground. As previously described, a significant amount of displacement current is induced as noise in a bit line that enters a positive sense current. The displacement current is caused by a change in the voltage between adjacent bit lines due to the mutual capacitance of the bit lines. Because of this, the bit line precharge period of the sensing cycle is sufficiently extended to wait until the displacement current is attenuated. During this period of delay, the memory binary fits into the bit line and thus draws additional power in the case where the bit line is pulled over the precharge circuit (4) the conduction current of the cell prevents it. By not implementing the bit line lock of the detected high current cell, only m is over-sensed and the bit line is not pulled to ground. The noise will be minimized and not 131985.doc •42- 200921679 Need to extend the bit line precharge cycle. This is schematically illustrated by the shortened precharge period 702 of the phase (2.5) in the Vt2 sensing cycle shown in FIG. In fact, it has been estimated that all of the unit's turn-on periods are shortened, resulting in power savings. In the event that the number of bit line locks is reduced in accordance with the present invention to improve sensing performance, more high current cells will remain in an on state to cause source bias errors as previously described in connection with Figures 12A and 12B. Regardless of the high level of "(10)R, a solution to minimize the error is to reference all the voltages reaching the control gates and drains of the individual memory cells with a node as close as possible to the source of the memory cell as a reference. For example, as shown in Figures 12A and i2B, the reference point can be selected at source line 32 instead of ground to minimize ground loop resistance. U.S. Patent Nos. 7,17,784 and 7,173,854 and Sekar et al., April 24, 2007, entitled "c〇mpensating SOURCE VOLTAGE DROP in NON-VOLATILE STORAGE" Techniques for minimizing source bias errors are disclosed in U.S. Patent Application Serial No. 7, the entire disclosure of each of which is incorporated herein by reference. The pre-equivalent potential is in the implementation of a sensing multi-state memory ("there is a relatively small sensing pass on the memory unit page relative to each other") By implementing a step of locking and identifying a bit line associated with a memory cell that is above a predetermined current level, bit line locking is used to reduce the benefits of total current and source bias errors and more The negative effect caused by the longer sensing time caused by the sub-cycle and the longer waiting time of the noise generated by the sub-circulation is balanced. By selecting 13I985.doc -43« 200921679 to reduce the number of locks, the reduction is reduced. Such negative effects of lower sensing performance and higher power consumption.Figure 17A illustrates an exemplary schedule for selectively enabling bit line locking in multiple passes of a multi-state sensing operation. Associated with the 8-state memory shown in Figure 8. To resolve all possible eight states, the memory cell pages will have to be sensed in at least 7 sensing cycles, each with a different demarcation threshold. Standard, such as Da*D b, ..., etc. Two pass sensing is implemented in each sensing cycle. However, the difference from the existing two pass sensing (see Figure 14) will disable the bit line locking operation during the second pass. In other words, every other sensing will be performed by skipping the bit line locking operation. In particular, the bit line locking operation will be skipped after sensing in the second pass. The number of passes and the existing two pass sensing At the same time, the bit line locking operation is reduced by 50%. In this case, less noise will be generated after the second pass and the first pass in the next cycle will be passed through with a shortened bit line. Charging cycle. If the word line settling time has the same magnitude as the bit line to bit 70 line crosstalk recovery number, then this embodiment does not significantly increase the performance, because the bit line lock and recovery from it can be compared with the next sensing bit. The steady word line voltages occur simultaneously. Figure 1 7B illustrates another exemplary schedule for selectively enabling bit line locks in multiple passes of a multi-state sensing operation. This example is also shown in Figure 8. 8 state memory is not relevant and similar The example shown in Figure 1 A 'except that two pass sensing will be performed in every other sensing sub-ring. In the cycle of solid pass, only the sensing operation will be performed and the pre-sensing will be skipped and Bit line locking operation. In this case, the number of passes is reduced by 131985.doc -44 - 200921679 5 0 and the number of bit line locking operations is reduced by 75. Figure 1 7C illustrates a store Pseudo-randomization of the second, page. In order to ensure a high current state in a page, the 己 八 布料 布料 布料 布料 布料 以 以 以 以 以 以 以 以 以 以 以 以 跨越 跨越 跨越 跨越 跨越 跨越 跨越 跨越 跨越 跨越 跨越 跨越 跨越 跨越 跨越 跨越 跨越In the body state, in the preferred embodiment 4: all possible records appear to be pseudo-randomly distributed in the possible memorization coding in the state, so as to be based on the skipped bit line locking operation. In the form of 10,000, the amount of current in the system is called _ # 忆. Estimated. t 〇己

—根據本發明之另一態樣’位元線鎖定之減小與記憶體單 疋頁t流動之不超過-預定最大電流之總電流相當。相反 地,在總電流(其係資料相依)看起來將要超過一預定之電 流位準時引用位元線鎖定操作。以此 八 在糸統不超 過一峰值電流之情形下最小化位元線鎖定操作。 預定之電流位準在一個感測一多狀態記憶體之實施方案 (其牽扯相對於多個狀態中之每一者對記憶體單元頁實施 多個感測通過)中,僅在該記憶體單元頁中流動的電流將 要超過預定最大電流時以預定感測通過實施鎖定與識別為 高於預定之電流位準之記憶體單元相關聯之位元線之步 驟。在一個實施例中,提供一電流監視器來監視記憶體單 元頁中流動之總電流。 圖1 8圖解說明回應於記憶體系統之一經監視之電流之位 元線鎖定操作。圖1 8圖解說明一正由類似於圖9中所示之 讀取/寫入電路270A中之對應感測模組感測之記憶體單元 頁中流動之電流。一電流監視器71 0定位於一聚集來自頁 131985.doc •45- 200921679 之個別記憶體單元之電流之總和之路徑處。舉例而言,將 其置於源極線34與系統接地之間之傳導路徑中。在一個實 細例中4電流監視器構建為一跨越一電阻器之電壓監視 器以使所監視之電流係由電壓降除以類似於圖中所 示之電阻給出。電流監視器71〇較佳包含邏輯以向狀態機 i i2(見圖丨)提供—信號BLNqL〇c。只要監視器電流^低 於-預先界定之位準,BLNqLC)c即輸出高。回應於 BLNoLOCW,該狀態機將致使在停用位元線鎖定之情 形下執行感測,同時控制信號GRS變為低(見圖1〇)。否 則在Zro:r>iA^z畤,BLN〇LOC將輸出低且此給該狀態機發 訊以在啓用位元線鎖定之情形τ操作感測以關閉某些記憶 體單元之電流。 圖19圖解說明其中根據已被鎖定之位元線之數目估計記 隐體單元頁中流動之總電流之另一實施例。圖1 9圖解說明 正由類似於圖9中所示之讀取/寫入電路27〇Α中之對應感 測模組感測之—記憶體單元頁中流動之電流。-累積器計 數已鎖存至接地之位元線之數目。接著,可將總電流^ 估計為每—單元之平均電流乘㈣頁中之未接地位元線之 數目。累積器720較佳包含邏輯以向狀態機112提供一信號 BLNoLOC。只要所估計之總電流卜⑽泡於一澇充芥定之位 準〜μ ’ BLNoLOC即輸出高。回應於BLN〇L〇c為高,該 狀態機將致使在停用位元線鎖定之情形下執行感測,同時 控制信號GRS變為低(見圖1〇)。否則,在心 BLNoLOC將輸出低且其傳訊狀態機以在啓用位元線鎖定 131985.doc -46 - 200921679 之晴形下操作感測以關閉某些記憶體單元之電流。 夕圖2 0圖解說明用於回應於一被超過之系統電流限制在— 多狀癌感消J操作之多通過中選擇性地啓用位元線鎖定之— 個例不性結果。與圖17A及i7B中所示之實例相反,位元 線鎖定操作係基於總電流(其係f料相依)之位準。因此, 舉二而言,在關於狀m則時引用兩通過感測以擺脫 最鬲電流狀態。接著針對狀態”B”在一個通過感測之情形- According to another aspect of the invention, the reduction in bit line lock is equivalent to the total current of the predetermined maximum current of the memory single page t flow. Conversely, the bit line locking operation is referenced when the total current (which is dependent on the data) appears to exceed a predetermined current level. In this case, the bit line locking operation is minimized in the case where the system does not exceed a peak current. The predetermined current level is in an embodiment of sensing a multi-state memory that involves performing multiple sensing passes on the memory cell page relative to each of the plurality of states, only in the memory cell The step in which the current flowing in the page will exceed a predetermined maximum current by predetermined sensing by performing a locking and associated bit line associated with a memory cell that is above a predetermined current level. In one embodiment, a current monitor is provided to monitor the total current flowing in the memory cell page. Figure 18 illustrates a bit line locking operation in response to a monitored current of one of the memory systems. Figure 18 illustrates a current flowing in a memory cell page that is being sensed by a corresponding sensing module in a read/write circuit 270A as shown in Figure 9. A current monitor 71 0 is positioned at a path that sums the sum of the currents of the individual memory cells from page 131985.doc •45-200921679. For example, it is placed in the conduction path between the source line 34 and the system ground. In a practical example, the 4 current monitor is constructed as a voltage monitor across a resistor such that the monitored current is divided by a voltage drop similar to the resistance shown in the figure. The current monitor 71A preferably includes logic to provide a signal BLNqL〇c to the state machine i i2 (see FIG. As long as the monitor current ^ is lower than the -predefined level, BLNqLC)c is the output high. In response to BLNoLOCW, the state machine will cause the sensing to be performed while the bit line is locked, while the control signal GRS goes low (see Figure 1). Otherwise, at Zro:r>iA^z畤, BLN〇LOC will output low and this will signal the state machine to operate sensing to disable the current of some memory cells in the case of enabling bit line lock. Figure 19 illustrates another embodiment in which the total current flowing in the page of the cryptographic unit is estimated based on the number of bit lines that have been locked. Figure 19 illustrates the current flowing in the memory cell page being sensed by a corresponding sensing module similar to the one in the read/write circuit 27A shown in Figure 9. - The number of bit lines that the accumulator count has latched to ground. Next, the total current ^ can be estimated as the number of ungrounded bit lines in the (4) page of the average current per unit. Accumulator 720 preferably includes logic to provide a signal BLNoLOC to state machine 112. As long as the estimated total current (10) is bubbled at a level of 芥~μ' BLNoLOC, the output is high. In response to BLN〇L〇c being high, the state machine will cause sensing to be performed with the bit line locked, while the control signal GRS goes low (see Figure 1〇). Otherwise, the heart BLNoLOC will output low and its signal state machine will operate to sense the current in some memory cells in the clear mode of the enable bit line lock 131985.doc -46 - 200921679. The eve diagram 20 illustrates an example result for selectively enabling bit line lock in response to an over-current system current limit in the multi-passion of the multi-cell cancer cancellation J operation. In contrast to the examples shown in Figures 17A and i7B, the bit line locking operation is based on the level of the total current (which is dependent). Therefore, for the second time, in the case of the shape m, two pass sensing is cited to get rid of the most current state. Then for the state "B" in a pass sensing situation

下V用位7L線鎖$。若此後,谓測到^篇要超過^,則 再人啓用^立TL線鎖定以針對狀態"c"實施兩通過感測。類 ㈣’若在狀期間位S線鎖定致使“減小,則將 再次啓用位元線鎖定。在狀態”D”及’Έ”處感測時此維持如 此,並允許皁通過感測直至感測"E"結束。在彼點處, 超過或將要超過“。因此,在㈣過中實施感測狀態 之下一循環等等。 ^ 圖2 1係-根據本發明之一較佳實施例圖解說明感測期間 之位元線鎖定控制之流程圖。 步称810 ·藉助相關聯之位元線及—共用字線達 組之個別記憶體單元之存取。 ^群 步称82G :選擇_將相對於其實施感測之分界臨限電壓 位準,該分界臨限電壓位準係一組多個臨限電壓位準中一 者。 步称83G:將該共用字線預充電至該選定分界臨限電壓 位準。 步称832 :將該等相關聯之位元線大致預充電至—預定 131985.doc -47- 200921679 電壓位準。 步驟840 :啓用位元線鎖定?若啓用,則進行至步驟 842’否則,進行至步驟85〇。 步驟842 :相對於該選定分界臨限電壓位準並行地感測 該記憶體單元群組。 〜 步称844 :識別感測到具有一小於該選定分界臨限電壓 位準之臨限電壓位準之任一記憶體單元。The lower V uses the 7L line lock $. If, after that, it is determined that ^^ exceeds ^, then the person activates the TL line lock to implement two-pass sensing for the state "c". Class (4) 'If the S-line lock causes "reduction" during the sizing period, the bit line lock will be enabled again. This is maintained during sensing at the states "D" and "Έ" and allows the soap to pass through the sense until sense Test "E" ends. At some point, it exceeds or will exceed ". Therefore, a cycle under the sensing state is implemented in (4), etc. ^ Figure 2 1 - illustrates the position during sensing according to a preferred embodiment of the present invention Flow chart of the line lock control. Step 810 · Access to the individual memory cells of the group by means of the associated bit line and the shared word line. ^ Group step 82G: Select _ will be sensed relative to it The boundary threshold voltage level is one of a plurality of threshold voltage levels. Step 83G: Precharge the common word line to the selected threshold threshold voltage level. Step 832: The pre-charged bit lines are pre-charged to a predetermined voltage level of 131985.doc -47- 200921679. Step 840: Enable bit line lock? If enabled, proceed to step 842' Otherwise, Go to step 85. Step 842: Sensing the memory cell group in parallel with respect to the selected threshold threshold voltage level. ~ Step 844: The identification sense has a threshold voltage less than the selected threshold. Any of the memory cells of the threshold voltage level.

步驟846 :藉由將任—所識別記憶體單元之相關聯之位 70線設定至一接地電位來鎖定該相關聯之位元線,針對小 於來自該組多個臨限電壓位準之電壓位準之所有疊代選擇 性地實施該鎖定。 ,騍850 :相對於該選定分界 該記憶體單元群組 ★步称860. i定之電壓等於該組中之最後電壓位準?若 等於’則進行至步驟870,否則返回至步驟82〇。 步称870 .完成對該群組之感測。 ”圖圖解說明其中執行實際感測而無任何預感測鎖定言 電流單元之一單通過感測之一替代實施例。回统 之電流負載位準或回應於-預定排程實施位元線鎖定'。、在先 與上升 讀敎及自其之恢復週期可 由傳導電流持續=二同時發生。以此模式,可藉 使此模式有效,需t :之:消增加之傳導電流。為 便可極快速^要字線啊間常數盡可能地減小,以 、、 個子線電壓位準至下一字線電壓位準進 131985.doc -48- 200921679 行感測。舉例而言,此 線來完成。可藉由立… 線或更多導電字 之狳栋祕 #由』充€所有位元線並在最後字線電屢 有位元線放電來最小化每一位元 快速感測所有壯能I、士 , n 導能量。 〜、、’減小消耗電流期間之時間來最小化傳 二實例係—具有16個狀態之記憶體。—個實施方案使 料2定在第3感狀狀態之後發生。所㈣—3感測操 、 、通,其中第3感測發現該等單元之4/16 =1/4接 通。基於此第三感測操作之結果,將立刻鎖定已發現在第 二感測操作期間被接通之所有單元。此外,位元線串擾恢 復次數將與自第三感測位準上升至第四感測位準之字線同 時發生、°下—位元'線鎖定操作發生在第7個感測位準處之 個込通感測與第8感測位準處之一個選通感測之間。下 -鎖定操作發生在㈣與^感測位準之間。如所解釋, 接著將僅存在3個敎㈣,域應H㈣擾恢復次數 中之每-次將與上升至下__位準之字線同時發生,而貫 整個事件序列,允許單元若且在其接通且在其到達下 定操作之前傳導DC電流。 圖23係一根據圖22中所示之替代實施例圖解說明感測期 間之位元線鎖定控制之流程圖。 、 步称910 :藉助相關聯之仇元線及一共用字線達成 組之個別記憶體單元之存取。 步称920 :選擇—將相對於其實施感測之分界臨限電壓 位準,該分界臨限電壓位準係一組多個臨限電壓位準中— 131985.doc -49- 200921679 者。 步驟930 :將該共用字線預充電至該選定分界臨限電壓 位準。 步驟932 :將該共用字線預充電至選定分界臨限電壓位 準。 步驟940 :相對於該選定分界臨限電壓位準並行地感測 該記憶體單元群組。Step 846: Lock the associated bit line by setting the associated bit 70 line of any identified memory cell to a ground potential for voltage bits less than the threshold voltage level from the set All of the iterations are selectively implemented for this lock. , 骒850: Relative to the selected boundary, the memory unit group ★ step 860. Is the voltage determined by the voltage equal to the last voltage level in the group? If it is equal to ', proceed to step 870, otherwise return to step 82. Step 870. Complete the sensing of the group. The figure illustrates an embodiment in which one of the pre-sensing lock current units is performed by one of the sense sensors without any pre-sensing lock current units. The current load level or the response to the predetermined schedule implementation bit line lock ' The first and the rising read period and the recovery period from it can be simultaneously caused by the conduction current continuous = two. In this mode, if this mode is valid, it needs t: it can increase the conduction current. ^ The constant between the word lines should be reduced as much as possible, and the voltage level of each sub-line to the next word line voltage level is 131985.doc -48- 200921679 line sensing. For example, this line is completed. You can use the line... or more conductive words to make all the bit lines and discharge the bit lines in the last word line to minimize each bit to quickly sense all the strong I , , n, energy. ~,, 'Reduce the period of current consumption to minimize the transmission of the second instance - memory with 16 states. - One embodiment makes the material 2 after the third state (4)—3 sense operation, and pass, of which the third sense is sent 4/16 of these units = 1/4 turn-on. Based on the result of this third sensing operation, all cells that have been found to have been turned on during the second sensing operation will be immediately locked. In addition, bit line crosstalk recovery The number of times will occur simultaneously with the word line rising from the third sensing level to the fourth sensing level, and the lower-bit 'line locking operation occurs at the seventh sensing level. Between a strobe sensing at the level of measurement. The lower-locking operation occurs between the (4) and ^ sensing levels. As explained, there will then be only 3 敎(4), and the domain should be H(4) for each of the recovery times - The second time will occur simultaneously with the word line rising to the next __ level, and the entire sequence of events allows the unit to conduct DC current if and before it is turned on and before it reaches the next operation. Figure 23 is based on Figure 22. An alternate embodiment of the illustration illustrates a flow chart of bit line lock control during sensing. Step 910: access to individual memory cells of the group by means of associated hatred lines and a common word line. 920: Selection—the threshold voltage level at which the sensing will be performed relative to the sensing. The threshold threshold voltage level is a set of multiple threshold voltage levels - 131985.doc -49- 200921679. Step 930: Precharge the common word line to the selected threshold threshold voltage level. : Precharging the common word line to a selected threshold threshold voltage level. Step 940: Sensing the memory cell group in parallel with respect to the selected threshold threshold voltage level.

步称950 :啓用位元線鎖定?若啓用,則進行至步驟 952,否則進行至步驟960。 步驟952 :識別感測到具有一小於該選定分界臨限電壓 位準之臨限電壓位準之任一記憶體單元。 步驟954 :藉由將任一所識別之記憶體單元之相關聯之 位元線,針對 之所有疊代選 位元線設定至一接地電位來鎖定該相關聯之 小於來自該組多個臨限電壓位準之電壓位準 擇性地實施該鎖定。 步驟960 :選定電壓等於該組中之最後電壓位準?若等 於’則進行至步驟970,否則返回至步驟92〇。 步驟970 :完成對該群組之感測。 圖24圖解說明其中執行實際感測而不鎖定任—古 阿%流單 元之一單通過感測之再一實施例。避免任—鎖定 \』彳糸一預 定排程之结果或可回應於該系統之電流負載位準。以此模 式,可藉由先前所述之位元線電壓之減小且亦減小之傳導 電流持續時間量來抵消增加之傳導電流。 如先前所述,兩個特徵幫助朝向實施減小感測通過之數 131985.doc -50- 200921679 目,甚至達到不需要兩通過感測之程度。—個係位元線電 壓之減小而另一係資料之隨機化或置亂。 極有效情形係,在讀取及驗證操作期間減小位元線電壓 以最小化傳導能量及電容性充電能量。減小之單元電流亦 將減輕CLSRC負載並允許較少的鎖定操作。感測電流— 亦必須減小以便在(VT,Iref)點處獲得一良好的跨導值,儘 官位70線電壓減小及所導致之接通電流減小。 然而,在習用記憶體架構中之感測方案之情形下,位元 線電壓之減小文到限制。此乃因習用感測方案採用位元線 電容來量測單元之傳導電流。纟元線帛充電至某一初始電 壓並接著允許由單元電流放電。纟電速率料元電流之一 量測。位元線電壓受到限制,乃因初始電壓不可如此低以 使其相對於位元線之R C時間常數放電至某一不可偵測之 值。另外,參考感測電流固定為位元線之RC常數且不可 輕易調節。 另一方面,在首先引入ABL記憶體架構中之感測方案 中,感測不相依於位元線之時間常數,乃因放電速率係相 對於卩有感測放大器之專用電容器。已在美國專利第 ,,93 1中揭示此等感測方案。可調節該專用電容器之 RC常數以最佳化感測。在此情形下,可進一步降低位元 線電屋J:匕一結果係將必須在一較低參考電流下實施感 測°此藉由電容n值之—適宜之選擇來輕易地完成。 ,資料置亂/隨機化之情形下,在每一狀態中存在粗略 相等數目之單π。意味著ln6的單元在該等狀態之每一 131985.doc •51 - 200921679Step 950: Enable bit line lock? If enabled, proceed to step 952, otherwise proceed to step 960. Step 952: Identifying any memory unit that senses a threshold voltage level that is less than the selected threshold threshold voltage level. Step 954: Locking the associated less than the threshold from the set by setting the associated bit line of any identified memory unit to all of the iterative selected bit lines to a ground potential The voltage level of the voltage level is selectively implemented to perform the locking. Step 960: Is the selected voltage equal to the last voltage level in the group? If it is equal to ', proceed to step 970, otherwise return to step 92. Step 970: Complete sensing of the group. Figure 24 illustrates yet another embodiment in which actual sensing is performed without locking one of the pass-through sensing. Avoid any - locking \" 彳糸 a predetermined scheduling result or can respond to the current load level of the system. In this mode, the increased conduction current can be offset by the decrease in the bit line voltage previously described and also the amount of conduction current duration that is reduced. As previously described, the two features help to reduce the number of passes through the implementation of 131985.doc -50- 200921679, even to the extent that no two-pass sensing is required. - The reduction of the line voltage of one line and the randomization or scrambling of another line of data. A very efficient case is to reduce the bit line voltage during read and verify operations to minimize conduction energy and capacitive charging energy. The reduced cell current will also reduce the CLSRC load and allow for less locking operation. The sense current—which must also be reduced to achieve a good transconductance at the (VT, Iref) point, is reduced by the 70-line voltage and the resulting turn-on current is reduced. However, in the case of a sensing scheme in a conventional memory architecture, the reduction in bit line voltage is limited. This is because the conventional sensing scheme uses bit line capacitance to measure the conduction current of the cell. The unit line is charged to an initial voltage and then allowed to discharge by the unit current. One of the current rate cell current measurements. The bit line voltage is limited because the initial voltage cannot be so low that it discharges to an undetectable value relative to the R C time constant of the bit line. In addition, the reference sense current is fixed to the RC constant of the bit line and cannot be easily adjusted. On the other hand, in the sensing scheme first introduced into the ABL memory architecture, the sensing does not depend on the time constant of the bit line because the discharge rate is relative to the dedicated capacitor of the sense amplifier. Such sensing schemes have been disclosed in U.S. Patent No. 1,93. The RC constant of the dedicated capacitor can be adjusted to optimize sensing. In this case, the bit line house can be further reduced. The result will be that the sensing will have to be performed at a lower reference current, which is easily accomplished by a suitable choice of the value of the capacitor n. In the case of data scrambling/randomization, there is a roughly equal number of single πs in each state. Means that the unit of ln6 is in each of these states. 131985.doc •51 - 200921679

者中。在資料經置亂之情形下,由於在每一較高控制閘極 電壓處僅1/1 6的單元接通,故較佳使用單通過(一個選通) 進行感測,且在其最初接通時,若狀態至狀態分開僅為 400 mV,則其不完全傳導接通。即使在先前所述之2通過 方案之情形下’將在第一選通之後鎖定之一顯著數目之單 元因在該第一選通期間存在之高單元源極負載而逃離鎖 定。該等逃離者將接著在位元線至位元線串擾恢復週期期 間傳導更多電流數量級,乃因已關閉未逃離者且已減小 CLSRC負載。故一由於其具有例如nA電流而逃離第一 選通之逃離者只要未逃離者被鎖定且CLSRC電壓減小即將 傳導700 nA。第一選通或第二選通整合時間為〇 4。 但,介入位元線至位元線串擾恢復次數為4 4 “Μ。在該 等條件下,因而2通過方案(亦即,兩個選通)在效能及能量 消耗兩個方面產生完全相反的結果。此在不使資料隨機 化’即/或在位元線電壓不可保持低,狀態至狀態分開係 實質(此非係8狀態、或16狀態、記憶體之情形)頗有用。即使在 具有較大狀態至狀態分開之4狀態記憶體中,#由僅減小 位元線電麼以使平均單元之接通電流為〜〇 3 UA,可使頁 上不止-半單元完全傳導且仍使用甚塊且更為能量有效之 本文所提及之所有專利、專财請案、論文、書籍、說 明書、其他出版物、文件及事物出於所有目的以::部内 容皆以引用方式併人本文中。在併人之出版物、文件或事 物中之任-者與本文件之文本之間之術語之界定或使用之 13I985.doc -52- 200921679 在本文件中術語的界定或使用將具 任何矛盾或衝突方面 有普遍性。 雖然已關於某些實施例描述本發明之各種態樣,然而應 \本發明有權在隨附申請專利範圍之之整個範鳴内受到 保護。 【圖式簡單說明】 圖1示意性地圖解說明其中可實施本發明之-非揮發性 s己憶體晶片之功能塊。 圖2示意性地圖解說明一非揮發性記憶體單元。 圖3針對浮動閘極可在任_時刻選擇性地儲存之四個不 同電荷Q1-Q4圖解說明源極_汲極電流1〇與控制閘極電壓 VCG之間的關係。 圖4圖解說明一 N〇R記憶體單元陣列之一實例。 圖5八不意性地圖解說明組織成一 nand串之一串記憶體 口 口 — 早兀。 圖5B圖解說明一由例如圖5A中所示之ΝΑΝ〇串構成之 NAND記憶體單元陣列之一實例。 圖6圖解說明用於藉助一系列交替程式化/驗證循環將一 記憶體單元頁程式化至一目標記憶體狀態之典型技術。 圖7(1)圖解說明一具有一經擦除狀態作為一接地狀態 "Gr”且漸進地更多經程式化之記憶體狀態"A"、"B”及"c” 之實例性4狀態記憶體陣列之臨限電壓分佈。 圖7(2)圖解說明一較佳、2位元LM編碼來代表圖7(1)中 所示之該四個可能的記憶體狀態。 131985.doc -53· 200921679 圖8 ( 1)圖解說明一實例性8狀態記憶體陣列之臨限電壓 分佈。 圖8(2)圖解說明一較佳' 3位元編碼來代表圖8(1)中所示 之該八個可能的記憶體狀態。 圖9圖解說明圖1中所示之跨越一記憶體單元陣列含有一 排感測模組之讀取/寫入電路。 圖1 〇更詳細地示意性地圖解說明適宜於實踐本發明之圖 9中所示之感測模組。Among them. In the case of scrambled data, since only 1/16 of the cells are turned on at each higher control gate voltage, it is better to use single pass (one gate) for sensing, and at the beginning of the connection In the meantime, if the state to state is separated by only 400 mV, it is not fully conductive. Even in the case of the previously described 2 pass scheme, a significant number of cells will be locked after the first gating to escape the lock due to the high cell source load present during the first gating. The evacuees will then conduct more current orders of magnitude during the bit-to-bit line crosstalk recovery period because the non-escaper has been turned off and the CLSRC load has been reduced. Therefore, the escaper who flees the first gate due to its current of, for example, nA, will conduct 700 nA as long as the non-escaper is locked and the CLSRC voltage is reduced. The first gating or second gating integration time is 〇 4. However, the number of crosstalk recovery from the bit line to the bit line is 4 4 "Μ. Under these conditions, the 2 pass scheme (ie, the two gates) produces the exact opposite in terms of efficiency and energy consumption. As a result, this does not randomize the data 'that is, or the bit line voltage cannot be kept low, and the state to state separation is substantial (this is not the state of 8 states, or the state of 16 states, memory). Even if it has In the state memory with a large state to a state separated, # is only reduced by the bit line so that the on-current of the averaging cell is ~〇3 UA, so that the page is more than - the half cell is completely conducted and still used. All patents, specialties, papers, books, brochures, other publications, documents, and articles mentioned in this article are for all purposes: for all purposes: Definition or use of terms between any of the publications, documents or things in the person and the text of this document 13I985.doc -52- 200921679 The definition or use of terms in this document will have any Contradictory or conflicting party While the various aspects of the invention have been described in terms of certain embodiments, the invention is intended to be protected within the scope of the appended claims. FIG. The functional map illustrates the functional blocks of the non-volatile simon memory wafer in which the present invention can be implemented. Figure 2 schematically illustrates a non-volatile memory cell. Figure 3 is selective for floating gates at any time. The four different charges Q1-Q4 stored in the ground illustrate the relationship between the source _ drain current 1 〇 and the control gate voltage VCG. Figure 4 illustrates an example of an N 〇 R memory cell array. The unintentional map illustrates the organization of a string of memory strings into a nand string - early. Figure 5B illustrates an example of a NAND memory cell array constructed of, for example, the string shown in Figure 5A. A typical technique for programming a memory cell page to a target memory state by means of a series of alternating stylization/verification cycles. Figure 7(1) illustrates an erased state as a ground state &q Uot; Gr" and progressively more routineized memory states "A", "B" and "c" an exemplary 4-state memory array threshold voltage distribution. Figure 7(2) illustrates a preferred, 2-bit LM code to represent the four possible memory states shown in Figure 7(1). 131985.doc -53· 200921679 Figure 8 (1) illustrates the threshold voltage distribution of an exemplary 8-state memory array. Figure 8(2) illustrates a preferred '3-bit encoding to represent the eight possible memory states shown in Figure 8(1). Figure 9 illustrates the read/write circuit of Figure 1 including a row of sensing modules across a memory cell array. Figure 1 is a schematic illustration of the sensing module shown in Figure 9 suitable for practicing the present invention in more detail.

圖11A更詳細地圖解說明圖丨〇中所示之預充電/钳位電 路。 圖11B更詳細地圖解說明圖丨〇中所示之單元電流鑑別器 電路。 圖12A圖解說明由於具有一有限對地電阻之源極線中之 電流而引起之源極電壓誤差問題。 圖12B圖解說明由一源極線電壓降所引起之記憶體單元 臨限電壓位準之誤差。 圖13(A)-13(J)係一藉助位亓&蚀~ )W 符1 70綠鎖定之2通過感測之定時 圖。 之8狀態記憶體上應 圖14示意性地圖解說明在圖8中所示 用現有兩通過感測方案之一實例。 圖應。 1 5圖解說明三個B比鄰位元線及其 之間之電容性耦合效 圖16(A)-16(J)係控制併入選擇性位 之操作之信號之定時圖。 元線鎖定之感測模 組 131985.doc 200921679 之多通過中選 之多通過中選 圖17Λ圖解說明用於在—多狀態感測操作 擇性地啓用位元線鎖定之一個例示性排程。 圖17B圖解說明用於在一多狀態感測操作 擇性地啓用位元線鎖定之另一例示性排程。 圖17C圖解說明 頁。 儲存僞IW機化之資料之記憶體單元 圖18圖解說明回應於記恃體糸餘夕 、。匕噶體糸統之一經監視之電流之位 元線鎖定操作。 圖19圖解說明其中藉助已祐错定 符初G被鎖疋之位兀線之數目估計記 憶體單元頁中流動之總電流之另一實施例。 圖20圖解說明用於回應於—被超過之系統電流限制在一 多狀態感測操#之多通過中選擇性地啓用位元線鎖定之一 個例示性結果。 圖21係一根據本發明之一較佳實施例圖解說明感測期間 之位元線鎖定控制之流程圖。 圖22圖解說明其中執行實際感測而無任何預感測鎖定高 電流單元之一單通過感測之一替代實施例。 圖23係一根據圖22中所示之替代實施例圖解說明感測期 間之位元線鎖定控制之流程圖。 圖24圖解說明其中執行實際感測而不鎖定任一高電流單 元之一單通過感測之再一實施例。 【主要元件符號說明】 10 記憶體單元 10-0 記憶體單元 131985.doc -55- 200921679 10-1 記憶體單元 10-2 記憶體單元 14 源極 16 汲極 20 電荷儲存單元 30 控制閘極 32 控制閘極 34 位元線 36 位元線 36-0 位元線 36-1 位元線 36-2 位元線 42 字線 44 選擇線 50 NAND 串 54 源極端子 56 汲極端子 100 記憶體晶片 110 控制電路 112 狀態機 200 二維記憶體單元陣列 230A 列解碼器 230B 列解碼器 231 資料I/O匯流排 131985.doc -56- 200921679 250A 頁多工器 250B 頁多工器 260A 行解碼器 260B 行解碼器 270A 讀取/寫入電路 270B 讀取/寫入電路 480 感測模組 481 感測節點 482 隔離電晶體 488 傳輸閘極 498 頁控制器 499 讀出匯流排 550 下拉電路/第一 η電晶體 552 第二η電晶體 600 感測放大器 610 \ 位元線電壓鉗位 J 612 電晶體 620丨 電壓鉗位 630 電晶體 631 節點 632 耦合電晶體 640 預充電/钳位電路 640' 預充電電路 647 節點 131985.doc -57- 200921679 650 651 652 656 660 710 720 Π 單元電流鑑別器 節點SEN 電容器 P通道電晶體 鎖存器 電流監視器 累積器Figure 11A illustrates in more detail the precharge/clamp circuit shown in Figure 。. Figure 11B illustrates in more detail the cell current discriminator circuit shown in Figure 。. Figure 12A illustrates the source voltage error problem due to current in the source line with a limited pair of ground resistance. Figure 12B illustrates the error in the memory cell threshold voltage level caused by a source line voltage drop. Fig. 13(A)-13(J) is a timing chart of 2 pass sensing by means of the position 亓 & ~~ )W symbol 1 70 green lock. 8 of the state memory should be schematically illustrated in Fig. 8 as an example of the existing two pass sensing scheme shown in Fig. 8. The picture should be. 1 5 illustrates the capacitive coupling effect of three B adjacent bit lines and between them. Figures 16(A)-16(J) are timing diagrams for controlling the signals incorporating the operation of the selective bits. The sensing mode set of the meta-line lock 131985.doc 200921679 The majority of the selection by the middle selection Figure 17A illustrates an exemplary scheduling for selectively enabling bit line locking in the multi-state sensing operation. Figure 17B illustrates another exemplary schedule for selectively enabling bit line locking in a multi-state sensing operation. Figure 17C illustrates the page. Memory Unit for Storing Pseudo IW Machined Data Figure 18 illustrates the response to the 恃 恃 糸 。. The bit line locking operation of one of the monitored currents of the corpus system. Figure 19 illustrates another embodiment in which the total current flowing in the page of the memory unit is estimated by the number of bit lines that have been locked by the initial error G. Figure 20 illustrates one exemplary result for selectively enabling bit line lock in response to a multi-state sensing operation in response to a system current limit being exceeded. Figure 21 is a flow chart illustrating bit line lock control during sensing in accordance with a preferred embodiment of the present invention. Figure 22 illustrates an alternate embodiment in which one of the single sensing passes is performed without any pre-sensing locked high current cells. Figure 23 is a flow diagram illustrating bit line lock control during sensing in accordance with an alternate embodiment shown in Figure 22. Figure 24 illustrates yet another embodiment in which actual sensing is performed without locking one of the high current cells. [Main component symbol description] 10 Memory unit 10-0 Memory unit 131985.doc -55- 200921679 10-1 Memory unit 10-2 Memory unit 14 Source 16 Deuterium 20 Charge storage unit 30 Control gate 32 Control gate 34 bit line 36 bit line 36-0 bit line 36-1 bit line 36-2 bit line 42 word line 44 select line 50 NAND string 54 source terminal 56 汲 terminal 100 memory chip 110 Control Circuit 112 State Machine 200 Two-Dimensional Memory Cell Array 230A Column Decoder 230B Column Decoder 231 Data I/O Bus 131985.doc -56- 200921679 250A Page Multiplexer 250B Page Multiplexer 260A Row Decoder 260B Row Decoder 270A Read/Write Circuit 270B Read/Write Circuit 480 Sensing Module 481 Sensing Node 482 Isolation Transistor 488 Transmission Gate 498 Page Controller 499 Read Bus 550 Pull Down Circuit / First η Transistor 552 second η transistor 600 sense amplifier 610 \ bit line voltage clamp J 612 transistor 620 丨 voltage clamp 630 transistor 631 node 632 coupling transistor 640 precharge / clamp Circuit 640 'node precharge circuit 647 131985.doc -57- 200921679 650 651 652 656 660 710 720 Π cell current discriminator capacitor node SEN P-channel latch transistor current monitor accumulator

U 131985.doc -58-U 131985.doc -58-

Claims (1)

200921679 十、申請專利範圍: 1. 一種非揮發性記憶體,其包括: 一記憶體單元陣列,其可藉由位元線及字線存取; 一感測電路群組,用於並行地感測一對應記憶體單元 群組中之傳導電流,其中該群組之個別記憶體單元可藉 由相關聯之位元線及一共用字線存取; 一字線電壓供應,用於相對於實施該感測之電壓位 準,將該共用字線預充電至一所選定之分界臨限電壓位 準,該分界臨限電壓位準係一組多個臨限電壓位準中之 一者; 一位元線電壓供應,用於將該相關聯之位元線大致預 充電至一預定電壓位準; 一組控制信號,用於相對於該選定之分界臨限電壓控 制忒感測電路群組,以並行地感測該記憶體單元群組;及 一位元線接地電路,用於在感測期間,使每一感測電 路回應於該每一感測電路之一感測結果及啓用位元線鎖 定之一控制信號兩者,將該相關聯之位元線接地。 2·如請求項1之非揮發性記憶體,其中: 該感測結果係該所感測之記憶體單元何時具有一高於 一參考感測電流之傳導電流;且 該控制信號係在滿足位元線鎖定之一啓用條件時被斷 定。 3,如凊求項2之非揮發性記憶體,其中: 該位元線鎖定啓用條件包含該記憶體單元群組之一聚 131985.doc 200921679 集電流何時達到一預定之電流位準。 4.如請求項2之非揮發性記憶體,其中: 該位元線鎖定啓用條件包含所識別之記憶 目何時達到一預定數目。 平几夂歎 5 ·如叫求項2之非揮發性記憶體,其中: 該位元線電壓供應在感測期間供應-預定之最小位元200921679 X. Patent Application Range: 1. A non-volatile memory comprising: a memory cell array accessible by bit lines and word lines; a sensing circuit group for parallel sense Detecting a conduction current in a group of corresponding memory cells, wherein individual memory cells of the group are accessible by associated bit lines and a common word line; a word line voltage supply for relative implementation The sensed voltage level pre-charges the common word line to a selected threshold threshold voltage level, the threshold threshold voltage level being one of a plurality of threshold voltage levels; a bit line voltage supply for substantially precharging the associated bit line to a predetermined voltage level; a set of control signals for controlling the sense circuit group relative to the selected threshold threshold voltage, Sensing the memory cell group in parallel; and a bit line grounding circuit for causing each sensing circuit to respond to one of the sensing circuits and enabling bits in each sensing circuit during sensing Line lock one of the control signals The bit line associated with the ground. 2. The non-volatile memory of claim 1, wherein: the sensing result is when the sensed memory cell has a conduction current higher than a reference sense current; and the control signal is in a bit One of the line locks is asserted when the condition is enabled. 3. The non-volatile memory of claim 2, wherein: the bit line lock enable condition comprises one of the memory cell groups. 131985.doc 200921679 When the current reaches a predetermined current level. 4. The non-volatile memory of claim 2, wherein: the bit line lock enable condition includes when the identified memory reaches a predetermined number. A few sighs 5 · The non-volatile memory of claim 2, where: the bit line voltage supply is supplied during sensing - the predetermined minimum bit 6.如請求項2之非揮發性記憶體,其中: 該位元線鎖定啓用條件包含該選定之分界臨限電壓位 準何時與該組多個臨限電壓位準之一 準一致 、、τ —預定位 7.如請求項6之非揮發性記憶體,其中: §亥子組係藉由在該組多個臨限電壓位準中選擇 距隔開的分界臨限電壓位準而形成。 ' 大致等 8·如請求項7之非揮發性記憶體,其中: 該記憶體單元群組儲存以一僞隨機圖案編瑪之 9.如請求項ό之非揮發性記憶體,其中: 该子組係一有序組之多個臨限電壓位準 界臨限彳立难· $ —、s 、隔η個分 介立旱之選擇,且η係一大於1的整數。 10.如請求項1之非揮發性記憶體,其中: s亥組分界臨限電壓含有至少三個電壓位準 11.如請求項1之非揮發性記憶體,其中: 該組分界臨限電壓含有至少七個電壓位準 12 ·如請求項1之非揮發性記憶體,其中: 131985.doc 200921679 該組分界臨限電壓含有至少十五個電壓位準。 13. 如請求項丨之非揮發性記憶體,其中該感測電路群組在 -讀取操作中操作,以讀取程式化至該記憶體單元群組 中之記憶體狀態。 、 14. 如吻求項丨之非揮發性記憶體’其中該感測電路群組在 一=式化操作之一部分中操作,以驗證是否已相對於該 h定之分界臨限電壓程式化該等記憶體單元中之任一 者。 15. 如請求項!之非揮發性記憶體,其中該非揮發性記憶體 單元群組係一快閃EEPROM之一部分。 〜 %如請求们之非揮發性記憶體,其中該快㈣咖 NAND類型。 系 17·如請求们之非揮發性記憶體,其中該群組之該等非揮 發性記憶體單4自具有至少—個電荷儲存元件。 18_如凊求項17之非揮發性記憶體,其t該電荷儲存元件係 一浮動閘極, 、 其中該電荷儲存元件係 其中該群組之該等非揮 19. 如請求項16之非揮發性記憶體 一介電層。 20. 如明求項丨之非揮發性記憶體 發性記憶體單元儲存至少兩個位元之資料 长項1至2G中之任—請求項之非揮發性記憶體,其 中該等非揮發性記憶體單元體現於一記憶體卡中。、 22. —種非揮發性記憶體,其包括: 一記憶體單元陣列,其可藉由位元線及字線存取; 13I985.doc 200921679 一感測電路群組,用於並行地感測一對應記憶體單元 群組中之傳導電流,其中該群組之個別記憶體單元可藉 由相關聯之位元線及一共用字線存取; 子線電壓供應,用於將該共用字線預充電至一將相 對於其實施該感測之選定之分界臨限電壓位準,該分界 臨限電壓位準係一組多個臨限電壓位準中之一者; 一位7L線電壓供應,用於將該相關聯之位元線大致預 充電至一預定電壓位準; 控制構件,用於相對於該選定之分界臨限電壓控制該 感測電路群組’以並行地感測該記憶體單元群組;及 一位元線接地電路,用於在感測期間,使每一感測電 =回應於該每-感測電路之—感測結果及啓用位元線鎖 心之一控制信號兩者,將該相關聯之位元線接地。 23. 如請求項22之非揮發性記憶體,其中該等非揮發性記憶 體單元體現於一記憶體卡中。 24. -種並行地感測—非揮發性記憶體單元群组之方法,直 包括: 〃 (a) 精由相關聯之付+括r u. ro ., 外1位70線及一共用字線,提供對該群 組之個別記憶體單元之存取; (b) 選擇一將相對於装眘<、B| 、 T 7、具實施該感測之分界臨限電壓位 準’該分界臨限電懕彳M z y 电竪位準係一組多個臨限電壓位準中之 _ _ m. · 9 (C)將該共用字線預充電1兮揖贪#人田> 几电主該選疋之分界臨限電壓位 131985.doc 200921679 (d)將該等相關聯之位元線大致預充電至一預定電壓 位準; (e)相對於該選定之分界臨限電壓位準,並行地感測 該記憶體單元群組; (0每當藉由將位元線接地而滿足一位元線鎖定啓用 條件時,即在繼'續進行至⑻之前實施⑻至㈣,否則 跳至(h):6. The non-volatile memory of claim 2, wherein: the bit line lock enable condition includes when the selected demarcation threshold voltage level is consistent with one of the plurality of threshold voltage levels of the group, τ - Pre-position 7. The non-volatile memory of claim 6, wherein: § hai sub-group is formed by selecting a spaced-off threshold threshold voltage level among the plurality of threshold voltage levels of the set. [Approximately: 8. The non-volatile memory of claim 7, wherein: the memory cell group is stored in a pseudo-random pattern. 9. The non-volatile memory of the request item, wherein: the sub- The group is an ordered group of multiple threshold voltages. The boundary is limited to the difficulty. $ —, s , and η are divided into the selection of the drought, and the η is an integer greater than 1. 10. The non-volatile memory of claim 1, wherein: the s-component threshold voltage comprises at least three voltage levels. 11. The non-volatile memory of claim 1, wherein: the component threshold voltage Containing at least seven voltage levels 12 • Non-volatile memory as claimed in claim 1, wherein: 131985.doc 200921679 The component threshold voltage contains at least fifteen voltage levels. 13. The non-volatile memory of claim 1, wherein the sensing circuit group operates in a read operation to read a memory state stylized into the memory cell group. 14. a non-volatile memory as in the case of a kiss, wherein the sensing circuit group operates in a portion of the operation to verify whether the threshold voltage has been programmed relative to the threshold Any of the memory units. 15. The non-volatile memory of claim 1 wherein the non-volatile memory cell group is part of a flash EEPROM. ~% like the non-volatile memory of the requester, which is the fast (four) coffee NAND type. The non-volatile memory of the group, wherein the non-volatile memory of the group has at least one charge storage element. 18_ The non-volatile memory of claim 17, wherein the charge storage element is a floating gate, wherein the charge storage element is the non-volatile portion of the group. Volatile memory - a dielectric layer. 20. The non-volatile memory memory unit of the present invention stores at least two bits of information in the length of 1 to 2G of the non-volatile memory of the request item, wherein the non-volatile memory The memory unit is embodied in a memory card. 22. A non-volatile memory comprising: a memory cell array accessible by bit lines and word lines; 13I985.doc 200921679 A sensing circuit group for parallel sensing a corresponding current in the group of memory cells, wherein the individual memory cells of the group are accessible by associated bit lines and a common word line; a sub-wire voltage supply for the common word line Precharging to a selected threshold voltage level relative to which the sensing is to be performed, the threshold threshold voltage level being one of a plurality of threshold voltage levels; a 7L line voltage supply Used to substantially precharge the associated bit line to a predetermined voltage level; a control component for controlling the sensing circuit group 'with respect to the selected threshold threshold voltage to sense the memory in parallel a unit cell group; and a one-element grounding circuit for controlling each sensing power in response to the sensing result of the per-sensing circuit and enabling one of the bit line locks during sensing Both of the signals ground the associated bit line. 23. The non-volatile memory of claim 22, wherein the non-volatile memory cells are embodied in a memory card. 24. A method of sensing in parallel—a group of non-volatile memory cells, including: 〃 (a) the associated payment + r u. ro ., the outer 1 bit 70 line and a common word a line that provides access to individual memory cells of the group; (b) selects a boundary that will be relative to the device, <, B|, T7, with the threshold voltage threshold for performing the sensing The threshold electric 懕彳 M zy electric vertical level is a set of multiple threshold voltage levels _ _ m. · 9 (C) pre-charging the common word line 1 兮揖 # 人 人 人 人 人 人 人 人 人 人The threshold of the threshold of the main election is 131985.doc 200921679 (d) The pre-charged bit lines are substantially precharged to a predetermined voltage level; (e) relative to the selected threshold threshold voltage level Sensing the memory cell group in parallel; (0) when a bit line lock enable condition is satisfied by grounding the bit line, that is, (8) to (4) are performed before continuing to (8), otherwise skipping To (h): (gl)識別經感測而具有—小於該選定之分界臨限電壓 位準之臨限電壓位準之任一記憶體單元; (g2)藉由將任一所識別却格牌留_ > ϋ , η。萌⑴屺隱體早兀之該相關聯之位元 線設定至一接地電位來銷定马如關 也,不頌疋β相關聯之位元線,每當滿 足一位元線鎖定啓用條件時, 臨限電壓位準之該等電壓位準 施該鎖定;及 即針對小於來自該組多個 之所有疊代,選擇性地實(gl) identifying any memory unit that is sensed to have a threshold voltage level that is less than the selected threshold threshold voltage level; (g2) by leaving any identified signature card _ > ϋ, η. Meng (1) 屺 hidden body early 兀 this associated bit line is set to a ground potential to pin Ma Ruguan, not 颂疋 β associated bit line, whenever a line lock enable condition is met The voltage level of the threshold voltage level is applied to the lock; and that is, for all generations less than the plurality of iterations from the group, selectively (h)針對來自該組多個臨限電麼位準之一下一電麼位 準,重複(b)至(h)直至已疊代來白 且代术自該組之每個電壓位 準。 25.如請求項24之方法,其中: 該實施(gl)至(g2)進一步包含: (g3)相對於該選定之分界臨限 該記憶體單元群組。 電壓位準,並行地感測 26.如請求項24之方法,其中: 集 該位元線鎖定啓用條件包含 你憶體早元群組之 電流何時達到一預定之電流位準。 聚 131985.doc 200921679 27. 如請求項24之方法,其中: 該位元線鎖定啓用條件包含所識別之記憶料元之 目何時達到一預定數目。 28. 如請求項24之方法,進一步包含: 在感測期間’以-預定之最小位元線電壓操作 元線。 29·如請求項24之方法,其中: 該位元線鎖定啓用條件包含該選定之分界臨限電塵位 準何時與該組多個臨限電壓位準之一子組中之—預定位 準一致。 3 0 ·如請求項2 9之方法,其中: 該子組係藉由在該la多個臨限電壓位準中選擇大致等 距隔開的分界臨限電壓位準而形成。 31·如請求項30之方法,其中: 該記憶體單元群組儲存以—棍p她圍电站 丨丁 Λ 偽Ik機圖案編碼之資料。 32. 如請求項29之方法,其中: 、 該子組係藉由在一有疼会日4 序、且之夕個臨限電壓位準中選 每隔η個分界臨限位準而形忐 〜攻且η係一大於1的整數。 33. 如請求項24之方法,其中: 該組分界臨限電壓含有?,丨 ,主J二個電壓位準。 34. 如請求項24之方法,其中: 該組分界臨限電壓含右5 ,丨、,, 3有至少七個電壓位準。 35·如請求項24之方法,其中: 該組分界臨限電壓含有、 啕至乂十五個電壓位準。 131985.doc 200921679 36. 如請求項24之方法,其中該感測係一讀取操作之一部 分,以讀取程式化至該記憶體單元群組中之記憶體狀 態。 37. ^請求項24之方法,其中該感測係一程式化操作之一部 分’以驗證是否已相對於該選定之/分界臨限電壓程式化 該等記憶體單元中之任一者。 38. 如請求項24之方法,其中該非揮發性記憶體單元群組係 一快閃EEPROM之一部分。 39. 如請求項24之方法,其中該快閃EEpR〇M係類 型。 4〇. ^請求項24之方法,其中該群組之該等非揮發性記憶體 單元各自具有至少一個電荷儲存元件。 41. 如請求項40之方法,其中該電荷儲存元件係一浮動閘 極0 42. 如請求項39之方法,其中該電荷儲存元件係一介電層。 』43.如請求項24之方法’其中該群組之該等非揮發性記憶體 單元儲存至少兩個位元之資料。 44.如請求項24至43中之任一請求項之方法,其中該等非揮 發性記憶體單元體現於一記憶體卡中。 131985.doc(h) Repeat (b) to (h) for each of the voltage levels from the set of multiple thresholds, repeating (b) to (h) until it has been repeated and from each voltage level of the group. 25. The method of claim 24, wherein: the implementations (gl) through (g2) further comprise: (g3) thresholding the group of memory cells relative to the selected boundary. Voltage level, sensing in parallel 26. The method of claim 24, wherein: the bit line lock enable condition includes when the current of the memory early group reaches a predetermined current level. 27. The method of claim 24, wherein: the bit line lock enable condition includes when the identified memory element reaches a predetermined number. 28. The method of claim 24, further comprising: operating the line at a predetermined minimum bit line voltage during the sensing period. The method of claim 24, wherein: the bit line lock enable condition includes when the selected demarcation threshold dust level is in a subset of the plurality of threshold voltage levels of the group - a predetermined level Consistent. The method of claim 29, wherein: the subgroup is formed by selecting a substantially equidistantly spaced boundary threshold voltage level among the plurality of threshold voltage levels. 31. The method of claim 30, wherein: the group of memory cells stores data encoded by the pseudo-Ik machine pattern of the power station. 32. The method of claim 29, wherein: the subgroup is shaped by selecting every n boundary thresholds in a viscous day 4 and at a threshold voltage level ~ Attack and η is an integer greater than one. 33. The method of claim 24, wherein: the component boundary threshold voltage is included? , 丨, the main J two voltage levels. 34. The method of claim 24, wherein: the component threshold voltage comprises a right 5, 丨,, 3 has at least seven voltage levels. 35. The method of claim 24, wherein: the component threshold voltage comprises, 啕 to fifteen voltage levels. The method of claim 24, wherein the sensing is a portion of a read operation to read a memory state stylized into the group of memory cells. The method of claim 24, wherein the sensing is part of a stylized operation to verify whether any of the memory cells have been programmed relative to the selected/demarcation threshold voltage. 38. The method of claim 24, wherein the non-volatile memory cell group is part of a flash EEPROM. 39. The method of claim 24, wherein the flash EEPR〇M is of a type. The method of claim 24, wherein the non-volatile memory cells of the group each have at least one charge storage element. The method of claim 40, wherein the charge storage element is a floating gate 0. 42. The method of claim 39, wherein the charge storage element is a dielectric layer. 43. The method of claim 24 wherein the non-volatile memory cells of the group store at least two bits of data. The method of any one of claims 24 to 43 wherein the non-volatile memory unit is embodied in a memory card. 131985.doc
TW97121282A 2007-06-07 2008-06-06 Non-volatile memory and method for improved sensing having bit-line lockout control TWI380307B (en)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
US11/759,898 US7492640B2 (en) 2007-06-07 2007-06-07 Sensing with bit-line lockout control in non-volatile memory
US11/759,909 US7489553B2 (en) 2007-06-07 2007-06-07 Non-volatile memory with improved sensing having bit-line lockout control

Publications (2)

Publication Number Publication Date
TW200921679A true TW200921679A (en) 2009-05-16
TWI380307B TWI380307B (en) 2012-12-21

Family

ID=39737581

Family Applications (1)

Application Number Title Priority Date Filing Date
TW97121282A TWI380307B (en) 2007-06-07 2008-06-06 Non-volatile memory and method for improved sensing having bit-line lockout control

Country Status (6)

Country Link
EP (1) EP2153443B1 (en)
JP (1) JP5016108B2 (en)
KR (1) KR101468883B1 (en)
CN (1) CN101711415B (en)
TW (1) TWI380307B (en)
WO (1) WO2008154229A1 (en)

Families Citing this family (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR101016078B1 (en) * 2009-01-21 2011-02-17 주식회사 하이닉스반도체 Nonvolatile Memory Device and Its Operation Method
JP2011258289A (en) * 2010-06-10 2011-12-22 Toshiba Corp Method for detecting threshold value of memory cell
WO2013147811A1 (en) * 2012-03-29 2013-10-03 Intel Corporation Method and system to obtain state confidence data using multistrobe read of a non-volatile memory
US9251903B2 (en) 2014-03-13 2016-02-02 Kabushiki Kaisha Toshiba Nonvolatile semiconductor memory device and control method thereof
JP6199838B2 (en) 2014-09-12 2017-09-20 東芝メモリ株式会社 Semiconductor memory device
JP2017059287A (en) * 2015-09-18 2017-03-23 株式会社東芝 Semiconductor device
US9881687B2 (en) * 2015-12-18 2018-01-30 Texas Instruments Incorporated Self-latch sense timing in a one-time-programmable memory architecture
CN110648714B (en) * 2018-06-26 2021-03-30 北京兆易创新科技股份有限公司 Data reading method and device, electronic equipment and storage medium
KR102797247B1 (en) * 2019-10-16 2025-04-21 에스케이하이닉스 주식회사 Memory device and operating method thereof
US12299562B2 (en) 2020-05-10 2025-05-13 Silicon Storage Technology, Inc. Analog neural memory array in artificial neural network with source line pulldown mechanism
US11430531B2 (en) * 2020-09-08 2022-08-30 Western Digital Technologies, Inc. Read integration time calibration for non-volatile storage

Family Cites Families (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7196931B2 (en) * 2002-09-24 2007-03-27 Sandisk Corporation Non-volatile memory and method with reduced source line bias errors
JP4420823B2 (en) * 2002-09-24 2010-02-24 サンディスク コーポレイション Nonvolatile memory and method with improved sensing behavior
US6956770B2 (en) * 2003-09-17 2005-10-18 Sandisk Corporation Non-volatile memory and method with bit line compensation dependent on neighboring operating modes
JP4287235B2 (en) * 2003-10-09 2009-07-01 株式会社東芝 Nonvolatile semiconductor memory device
US7251160B2 (en) * 2005-03-16 2007-07-31 Sandisk Corporation Non-volatile memory and method with power-saving read and program-verify operations
US7173854B2 (en) * 2005-04-01 2007-02-06 Sandisk Corporation Non-volatile memory and method with compensation for source line bias errors

Also Published As

Publication number Publication date
TWI380307B (en) 2012-12-21
KR101468883B1 (en) 2014-12-04
KR20100029192A (en) 2010-03-16
EP2153443A1 (en) 2010-02-17
CN101711415A (en) 2010-05-19
JP2010529585A (en) 2010-08-26
WO2008154229A1 (en) 2008-12-18
CN101711415B (en) 2014-10-22
JP5016108B2 (en) 2012-09-05
EP2153443B1 (en) 2013-02-20

Similar Documents

Publication Publication Date Title
TW200921679A (en) Non-volatile memory and method for improved sensing having bit-line lockout control
CN101946287B (en) Low noise sense amplifier array and method for nonvolatile memory
JP4420823B2 (en) Nonvolatile memory and method with improved sensing behavior
CN102113057B (en) High speed sense amplifier array and method for nonvolatile memory
KR101069750B1 (en) Memory sensing circuit and method for low voltage operation
US8531889B2 (en) Non-volatile memory and method with improved sensing having bit-line lockout control
JP4988156B2 (en) Nonvolatile memory and method with reduced adjacent field errors
JP4898117B2 (en) Nonvolatile memory and method with reduced bitline crosstalk errors
US7428171B2 (en) Non-volatile memory and method with improved sensing
TWI380308B (en) Read operation for non-volatile storage that includes compensation for coupling
TW201203259A (en) Mitigating channel coupling effects during sensing of non-volatile storage elements
US20080304325A1 (en) Non-volatile memory with improved sensing having bit-line lockout control
US20130100744A1 (en) Compact Sense Amplifier for Non-Volatile Memory

Legal Events

Date Code Title Description
MM4A Annulment or lapse of patent due to non-payment of fees